CN101023867A - Data correction apparatus, data correction method, magnetic resonance imaging apparatus and X-ray CT apparatus - Google Patents

Abstract

A data correction apparatus includes a sensitivity correction unit and an SNR distribution correcting unit. The sensitivity correction unit produces first processed data by performing sensitivity correction to first objective data obtained based on correction objective data using ununiform sensitivity distribution of a sensor for acquiring the correction objective data. The SNR distribution correcting unit produces pieces of component data each subjected to corresponding weighting depending on an SNR distribution and corresponding filtering having a mutually different intensity using second objective data obtained based on the correction objective data to produce second processed data by compounding the pieces of the component data.

Description

Data correction apparatus, data correcting method, MR imaging apparatus and X ray CT device

Technical field

The present invention relates to data correction apparatus, data correcting method, MR imaging apparatus and X ray CT (the computer X-ray chromatography is according to the shadow art) device, they are to because the sensitivity table of the pick off that display space distributes reveals the data that the non-homogeneous SNR in space (signal to noise ratio) distributes proofreaies and correct, so that be uniform through gauged data.

Background technology

MRI (nuclear magnetic resonance) device is used as monitoring arrangement (for example referring to Japan Patent No.3135592) at medical domain traditionally.

The MRI device be by the gradient coil in the imaging region that is placed on the cylindrical static field magnet object that is used for producing magnetostatic field generate gradient magnetic, by send from the RF coil RF (radio frequency) signal make nuclear spin generation magnetic resonance the object, and by using because NMR (nuclear magnetic resonance, NMR) signal of excitation generation is rebuild the device of the image of object.

In MRI device in recent years, in order to be accelerated into picture, the RF coil is made of the integral body that is used to send (WB) coil and the phased-array coil that is used to receive.Phased-array coil comprises a plurality of surface coils, thereby might reduce imaging time, because each surface coils receives the NMR signal simultaneously, obtains more data in short interval.

Yet when the RF coil was made of phased-array coil and WB coil, the signal intensity of the view data that obtains by reconstruction process and NMR signal was because the heterogeneity of the sensitivity of phased-array coil or WB coil also has heterogeneity.Usually, the heterogeneity of the sensitivity of WB coil is enough little, is in insignificant level.Yet the heterogeneity of the sensitivity of each surface coils in the phased-array coil that uses for every kind of purposes is very big, and influences view data.

For this reason, must to since the heterogeneity of the signal intensity of the view data that the heterogeneity of the sensitivity of phased-array coil causes proofread and correct.

In view of foregoing, up to now, before the main scanning of the image that is used to generate object, carry out the sensitivity prescan.Then, by the sensitivity prescan, obtain view data from each phased-array coil and WB coil.According to signal intensity rate, i.e. the division numerical value of the signal intensity of image data section, the sensitivity profile of phased-array coil is estimated as three-dimensional sensitivity map datum.And by using the three-dimensional sensitivity map datum of the phased-array coil that obtains like this, the signal intensity heterogeneity of view data is corrected.

Yet the MR imaging of a plurality of surface coils of stating in the use and using in the MR imaging of single surface coils when the sensitivity profile of correction surface coil, has a problem: the space heterogeneity promptly occurs in SNR.That is, before proofreading and correct, the sensitivity profile of surface coils is that the space is heterogeneous, but the picture noise level is constant.

Therefore, if the sensitivity profile of surface coils is corrected, and the signal intensity of view data is set to constantly according to the space, and then to become the space heterogeneous for picture noise.For example, has bigger intensity at the picture noise at the part place that the correction of signal intensity by sensitivity profile is exaggerated compared with the picture noise at the part place that is not reinforced at signal intensity.As a result, SNR becomes spatial inhomogeneity distribution, and this causes deterioration of image, so the space heterogeneity of SNR is undesirable when diagnosis.

In addition, at the medical apparatus such or be different from and use in the biological information measuring instrument of surface coils as the MRI device of pick off such as other image diagnosing system, if the intensity of the signal that obtains when the space of collecting sensor non-uniform sensitivity distributes is set to constant, then occur the space heterogeneity in SNR and noise, this can cause the deterioration of picture quality or measurement result.

Summary of the invention

The present invention is that the situation of considering routine is made, the purpose of this invention is to provide a kind of data correction apparatus, data correcting method, MR imaging apparatus and X ray CT device, the spatially uniform that they can keep SNR to distribute by distributing with the simple space non-uniform sensitivity of handling correcting sensor simultaneously, thus uniform data obtained.

In aspect of this target of realization, the invention provides a kind of data correction apparatus, comprise: the sensitivity correction unit is configured to distribute by the non-uniform sensitivity that use is used to obtain the pick off of correction target data and carries out sensitivity correction for first target data that obtains according to the correction target data and produce first deal with data; With the SNR distribution correcting unit, be configured to produce the component data section, each component data section distributes according to SNR and is subjected to respective weight, and the corresponding filtering by using second target data that obtains according to the correction target data to have different mutually intensity, so that produce second deal with data by mixing these component data sections.

In aspect of this target of realization, the present invention also provides a kind of data correcting method, may further comprise the steps: the sensitivity profile heterogeneous that is used to obtain the pick off of correction target data by use is carried out sensitivity correction for first target data that obtains according to the correction target data and is produced first deal with data; And produce into the component data segment, each component data section is subjected to depending on the corresponding weighting of SNR distribution, and the corresponding filtering by using second target data that obtains according to the correction target data to have different mutually intensity, so that produce second deal with data by mixing these component data sections.

In aspect of this target of realization, the present invention also provides a kind of MR imaging apparatus, comprising: coil; Data capture unit is configured to obtain the magnetic resonance image data of object and at least one item in the k spatial data by the described coil that is used as pick off; The sensitivity correction unit is configured to carry out sensitivity correction by first target data that obtains according at least one item in magnetic resonance image data and the k spatial data for the sensitivity profile heterogeneous of using described coil and produces first deal with data; And SNR distribution correcting unit, be configured to produce the component data section, each component data section distributes according to SNR and is subjected to respective weight, and the corresponding filtering by using second target data that obtains according to magnetic resonance image data and at least one item in the k spatial data to have different mutually intensity, so that produce second deal with data by the mixed components data segment.

In aspect of this target of realization, the present invention also provides a kind of X ray CT device, comprising: X-ray detector; Data capture unit is configured to obtain the view data of object and at least one item in the data for projection by the described X-ray detector that is used as pick off; The sensitivity correction unit is configured to carry out sensitivity correction by first target data that obtains according at least one item in view data and the data for projection for the sensitivity profile heterogeneous of using described X-ray detector and produces first deal with data; And SNR distribution correcting unit, be configured to produce the component data section, each component data section distributes according to SNR and is subjected to respective weight, and the corresponding filtering by using second target data that obtains according to view data and at least one item in the data for projection to have different mutually intensity, so that produce second deal with data by the mixed components data segment.

The spatially uniform that aforesaid data correction apparatus, data correcting method, MR imaging apparatus and X ray CT device can keep SNR to distribute by distributing with the simple space non-uniform sensitivity of handling correcting sensor simultaneously, thus uniform data obtained.

Description of drawings

In the accompanying drawings:

Fig. 1 is the functional block diagram that shows according to the data correction apparatus of the first embodiment of the present invention;

Fig. 2 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus shown in Figure 1 for the view data execution sensitivity correction of obtaining from image diagnosing system simultaneously;

Fig. 3 is the functional block diagram that shows data correction apparatus according to a second embodiment of the present invention;

Fig. 4 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus shown in Figure 3 for the view data execution sensitivity correction of obtaining from image diagnosing system simultaneously;

Fig. 5 is the functional block diagram that shows the data correction apparatus of a third embodiment in accordance with the invention;

Fig. 6 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus shown in Figure 5 for the view data execution sensitivity correction of obtaining from image diagnosing system simultaneously;

Fig. 7 is the functional block diagram that shows the data correction apparatus of a fourth embodiment in accordance with the invention;

Fig. 8 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus shown in Figure 7 for the view data execution sensitivity correction of obtaining from image diagnosing system simultaneously;

Fig. 9 is the structure chart that shows MR imaging apparatus according to an embodiment of the invention;

Figure 10 is the figure of demonstration by the example of the detailed structure of RF coil shown in Figure 9;

Figure 11 is the sectional view that shows the exemplary arrangement of WB coil shown in Figure 10 and phased-array coil;

Figure 12 is the functional block diagram that shows computer shown in Figure 1;

Figure 13 shows by MR imaging apparatus shown in Figure 9 to obtain the image of object and keep the flow chart of the process of SNR distributing homogeneity subsequently for the image execution of obtaining about the sensitivity correction of each surface coils simultaneously;

Figure 14 shows the ideal abdomen images S after the sensitivity correction of being supposed by data correction apparatus shown in Figure 1 in the emulation of image rectification _{Ideal_scor}

Figure 15 shows the original image S before the sensitivity correction of being used by data correction apparatus shown in Figure 1 in the emulation of image rectification _Orig

Figure 16 display sensitivity distribution IS _Sens, it is used to the original image S shown in Figure 15 for the coil of abdominal part and profile thereof _OrigCarry out sensitivity correction;

Figure 17 demonstration is passed through for the coil of abdominal part and the original image S shown in Figure 15 of profile thereof _OrigThe abdomen images S that carries out sensitivity correction and obtain _{Ideal_scor}

Figure 18 display noise distribution noise_scor, it is used to be undertaken by data correction apparatus shown in Figure 1 the emulation of image rectification after sensitivity correction;

Figure 19 demonstration is passed through with even LSI wave filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile carries out that SNR proofreaies and correct and the image that obtains;

Figure 20 shows by using the LSI wave filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile is carried out the SNR nonuniformity correction of the addition be attended by weighting and the image that obtains;

Figure 21 shows by using the homogeneous texture sef-adapting filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile carries out that SNR proofreaies and correct and the image that obtains;

Figure 22 shows by using the homogeneous texture sef-adapting filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile is carried out the SNR nonuniformity correction of the addition be attended by weighting and the image that obtains;

Figure 23 carries out under the situation of filtering operation the figure of the standard deviation of noise and the variation of RMSE by changing level and smooth intensity when being presented at by the gauged emulation of data correction apparatus carries out image shown in Figure 1;

Figure 24 be illustrated in by data correction apparatus carry out after the sensitivity correction and non-homogeneous filtering before the non-uniform Distribution of noise and the concept map of standard deviation;

Figure 25 is that expression is by carrying out the distribution of the noise that homogenization obtains and the concept map of standard deviation with non-homogeneous filtering for the non-uniform Distribution of noise shown in Figure 24;

Figure 26 is that the filter function that is presented at the even wave filter in the data correction apparatus is the figure of the example under the situation of Hanning function;

Figure 27 is the even flow chart of the processing stream under the situation of the filtering of level and smooth intensity the best of wave filter in the filter cell that is presented in the data correction apparatus shown in Figure 1;

Figure 28 is the structure chart that shows X ray CT device according to an embodiment of the invention; And

Figure 29 be illustrated in the imaging region of X ray CT device shown in Figure 28 the position and from the figure of the relation between the intensity of the x-ray detection signal of each X-ray detector output.

The specific embodiment

Referring now to accompanying drawing data correction apparatus, data correcting method, MR imaging apparatus and X ray CT device are according to an embodiment of the invention described.

Fig. 1 is the functional block diagram that shows according to the data correction apparatus of the first embodiment of the present invention.

Data correction apparatus 1 is made of the computer of fetch program.Should be pointed out that data correction apparatus 1 is whole or a part ofly can constitute by circuit.Data correction apparatus 1 comprises sensitivity correction unit 2, SNR distribution acquiring unit 3, filter cell 4 and weighting summation unit 5.Therefore, these unit are by for by such as the medical apparatus of image diagnosing system be used for data that the measuring device of biological information obtains and carry out the spatially uniform that correction that the space non-uniform sensitivity of pick off distributes is handled and kept SNR to distribute simultaneously, thereby provide the function of obtaining even data for data correction apparatus 1.

The example that is used to collect as the medical apparatus of the data of correction target comprises biological information measuring instrument, such as electroencephalogram, electrocardiogram and oscillosynchroscope, and image diagnosing system, such as supersonic diagnostic appts, radiodiagnosis device, X ray CT device, MR imaging apparatus or nuclear medicine diagnostic apparatus.The pick off of MR imaging apparatus is a coil.Radiodiagnosis device, X ray CT device and be a detecting unit such as each pick off of the nuclear medicine diagnostic apparatus of SPECT (monochromatic light gives chromatographical X-ray that radiation calculates according to the shadow art) or PET (chromatographical X-ray that positron emission calculates is according to the shadow art).The detecting unit of radiodiagnosis device, X ray CT device or nuclear medicine diagnostic apparatus has the type that comprises that Direct Transform type and indirect variable are remodeled.For any type, detecting unit has sensitivity heterogeneous, needs to carry out sensitivity correction.The pick off of supersonic diagnostic appts is the probe that is equipped with a plurality of ultrasonic transducers.

According to data correction apparatus 1, be not only the view data of collecting, and also have because the non-uniform sensitivity distribution of pick off needs the various data of sensitivity correction by image diagnosing system, all can be set to correction target.For example, when the medical apparatus that is used to collect the correction target data is MR imaging apparatus, MR view data not only, and also have the k spatial data all can be set to correction target.In addition, when the medical science that is used to collect the correction target data is established when being X ray CT device, X ray CT view data not only, and also have data for projection all can be set to correction target.

Then, the data as the correction target of data correction apparatus 1 can be set to the data of dimension arbitrarily.For example, one-dimensional data, 2-D data, three-dimensional data or represent the locus and the four-dimensional data of time can be set to the correction target of data correction apparatus 1.Example with data of time dimension is included in the data with time shaft that obtain in electroencephalogram, electrocardiogram, oscillosynchroscope or the supersonic diagnostic appts.In addition, the view data of view data by using the T1 weighting that T1 relaxation (longitudinal relaxation) time difference obtains and the T2 weighting that obtains by use T2 relaxation (transverse relaxation) time difference also is the example with data of time dimension in the MRI device.The view data of the view data of T1 weighting and T2 weighting disappears in time and decays.If carry out the sensitivity correction of coil, then noise becomes heterogeneous in time.For this reason, also need carry out correction to data at time-axis direction handles.

After this, are explanations of the example of the view data of collection in image diagnosing system 6 with the data that provide as correction target.Image diagnosing system 6 comprises pick off 7, image data acquisition unit 8, image data memory cell 9, sensitivity map memory element 10 and display unit 11.Pick off 7 is configured to detect data under the control by image data acquisition unit 8, and the data that detect are provided to image data acquisition unit 8.

Image data acquisition unit 8 is equipped with by controlling the function that pick off 7 is collected data and generated view data from the data of collecting.Image data memory cell 9 comprises the function of storage by the view data of image data acquisition unit 8 generations.Sensitivity map memory element 10 has the space of storage representative sensor 7 or the function of the sensitivity map that time sensitivity distributes.Display unit 11 comprises display, and has on display the function that shows the view data of reading from image data memory cell 9.

The sensitivity map that is stored in the sensitivity map memory element 10 can obtain by estimating or measuring according to any means.Especially, by carrying out data collection, also can generate the sensitivity map according to the data of collecting for the sensitivity geographic survey value of using pick off 7.

The sensitivity correction unit 2 of data correction apparatus 1 has from image data memory cell 9 and obtains function as the raw image data of the target of sensitivity correction, also have from sensitivity map memory element 10 and obtain the sensitivity map that will be used in sensitivity correction, so that when using sensitivity map memory element, by raw image data is carried out the function that sensitivity correction obtains the sensitivity correction view data, and have the function that the sensitivity correction view data that obtains is like this offered filter cell 4 and weighting summation unit 5.

SNR distribution acquiring unit 3 has obtaining or estimates distribution together with the SNR that generates for the sensitivity correction of raw image data according to any means, so that the distributed intelligence of relevant SNR is provided to weighting summation unit 5.The distributed intelligence of relevant SNR for example can be set to represent the SNR distribution window of the distribution of SNR.The distributed intelligence of relevant SNR can be estimated from the sensitivity map being stored in sensitivity map memory element 10.In view of above content, SNR distribution acquiring unit 3 has the function of obtaining the sensitivity map from sensitivity map memory element 10.

In addition, also can be calculated about the distributed intelligence of SNR by carrying out such as low-pass filtering and the such various Flame Image Process of threshold process as the sensitivity correction view data of the correction target of non-uniform Distribution SNR with respect to raw image data or after sensitivity correction.In addition, as another example,, also can obtain the distributed intelligence of relevant SNR by comparing as the raw image data of correction target view data with the image that obtains dividually.So SNR distribution acquiring unit 3 also can have the image processing function of this distributed intelligence that is used to calculate relevant SNR.Except this function, SNR distribution acquiring unit 3 can have the function of the distribution of the function of the distribution of measuring SNR or the SNR that input had before been measured.

Filter cell 4 has by using even wave filter that the sensitivity correction data that 2 sensitivity correction view data that receive or the conversion by the sensitivity correction view data from the sensitivity correction unit obtain are carried out the function that Filtering Processing generates filtered view data or filtering data, and the function that the filtered view data that the filtered view data that generates like this or the conversion by filtering data obtain is offered weighting summation unit 5.Even wave filter with arbitrary number of different mutually filtering strengths is provided for filter cell 4 where necessary.Then, filter cell 4 is configured to generate and is subjected to based on the single filtered image data of the Filtering Processing of even wave filter or is subjected to multistage filtered image data based on the Filtering Processing of the even filtering with different mutually filtering strengths.

Here, be subjected to sensitivity correction as the sensitivity correction view data of Filtering Processing target, thus SNR be spatially or the time go up heterogeneous.Yet evenly wave filter can be formed by being applied to the general general filter that its SNR is assumed to be constant data.That is, can be looked at as on big meaning in the space and/or on the time be that uniform nearly all wave filter can be employed as even wave filter to its characteristic.For example, each has even core (filtering strength), have in time with the space on the linear filter of constant intensity, perhaps its core can constitute even wave filter according to the structure adaptive mode filter that data structure is determined.

In addition, the data as the Filtering Processing target can obtain in the MRI device, the real space data on r space (real space) or the k spatial data on the k space.When the k spatial data was set to the Filtering Processing target, the sensitivity correction view data was transformed into sensitivity correction k spatial data by FT (Fourier transform), and sensitivity correction k spatial data becomes the Filtering Processing target.Then, Filtering Processing k data are transformed into the filtered view data that will be provided for weighting summation unit 5 by FT after Filtering Processing.

And, proposed its midband by use divided specific FREBAS (the Frenel transform band is cut apart) space of Frenel conversion and wherein filter strength be confirmed as making that SNR becomes SNR self-adaptation type Wiener wave filter best in handling the space.The Wiener wave filter can be provided with by being different from the spatial wavelet transformation division fourier space of FREBAS or the real space for handling the space that the space obtains.Especially, if the FREBAS space is set to the processing space of Wiener wave filter, the core that then is added to the even wave filter of the spatial data of FREBAS can be appropriately determin by the supervision noise,

In view of foregoing, the sensitivity correction view data is transformed into sensitivity correction FREBAS spatial data, be subjected to being transformed into filtered view data based on the Filtering Processing FREBAS spatial data of the Filtering Processing of Wiener wave filter, it can be provided to weighting summation unit 5.The FREBAS space is the space that is used to based on the analysis of multiresolution analysis method, this multiresolution analysis method is used the Frenel conversion a plurality of separate or band segmentation is improved one's methods as one that is used for SNR.

The details that should be pointed out that structure adaptive filter is at Chen, H.G., A.Li, L.Kaufman, and J.Hale, describe among " the A fast filtering algorithm for imageenhancement ", IEEE Trans.Medical Imaging 13 (3): 557-564 (1994).The details of Wiener wave filter is at Ito S, Yamada Y, " Use of Dual FresnelTransform Pairs to Improve Signal-to-Noise Ratio in MagneticResonance Imaging " describes among the Med Imag Tech 19 (5), 355-368 (2001).

Weighting summation unit 5 have for the single or multiple filtered image data section that receives from filter cell and before filtering is processed from the sensitivity correction unit the 2 sensitivity correction view data that receive according to such as from SNR distributed acquisition unit the such SNR distributed intelligence of the 3 SNR distribution window that receive carry out weighting summation, to generate the function that basically is equivalent to the filtered view data of heterogeneity that is subject to the view data processed based on the filtering of non-homogeneous wave filter and the function that the filtered view data of heterogeneity that generates like this is written to the image data memory cell 9 of image diagnosing system 6. Should be pointed out that the sensitivity correction view data before Filtering Processing is not set to the target of weighting summation, and have only the target that can be set to weighting summation from a plurality of filtered image data section of filter cell 4 receptions.In other words, be subjected to having different intensity filtering, a plurality of filtered image data sections can be set to the target of weighting summation.

Then, the operation of data correction apparatus 1 and the explanation of action will be provided.

Fig. 2 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus shown in Figure 11 for the view data execution sensitivity correction of obtaining from image diagnosing system 6 simultaneously.Each step that in Fig. 2, has the symbolic representation flow chart of S and numeral.

At first, view data is collected in the image diagnosing system 6 in advance.That is, according to the control from image data acquisition unit 8, the data that are used for the image generation are detected by pick off 7.The data that detect are provided to image data acquisition unit 8 from pick off 7, and image data acquisition unit 8 generates view data from these data.Then, the view data that is generated is written to image data memory cell 9, and as raw image data S _OrigBe stored.Simultaneously, estimate according to any means or the time and/or the spatial sensitivity profile of measuring transducer 7.The sensitivity profile of the pick off 7 that obtains like this is written to sensitivity map memory element 10, and as sensitivity map I _SensBe stored.

Then, at step S1, SNR distribution acquiring unit 3 obtains the sensitivity map I that is used in sensitivity correction from sensitivity map memory element 10 _Sens, and according to sensitivity map I _SensObtain the distributed intelligence of relevant SNR.The distributed intelligence of relevant SNR is set to weighting function W _Snr, as the SNR distribution window, and weighting function W _SnrBe provided to weighting summation unit 5.

If sensitivity map I _SensSpatially the three dimensional constitution with x direction, y direction and z direction distributes, then sensitivity map I _SensCan be by normalization, and be represented as I _Sens(x, y, z).At image diagnosing system 6 is under the situation of MRI device, sensitivity map I _Sens(x, y z) become the real space (r space) data of representing coil sensitivity profiles.

Represent the weighting function W of the distribution of SNR _Snr(x, y z) can be by the normalization sensitivity map I of the whole bag of tricks from pick off 7 _SensObtain.For example, having only the sensitivity map I of pick off 7 _Sens(x, y, maximum max[I z) _Sens(x, y, z)] by standardization and weighting function W _Snr(x, y, maximum z) are set under 1 the situation, weighting function W _Snr(x, y z) can be determined according to formula (1).

W _snr(x，y，z)＝I _sens(x，y，z?)/max[I _sens(x，y，z)] (1)

W wherein _Snr(x, y, z): weighting function (SNR distribution function).

In addition, for example, at weighting function W _Snr(maximum z) is set to 1 for x, y, and weighting function W _Snr(minima z) is set to also use the sensitivity map I of pick off 7 under 0 the situation for x, y _Sens(x, y, minima min[I z) _Sens(x, y, z)], can determine weighting function W according to formula (2) thus _Snr(x, y, z).

W _snr(x，y，z)

＝{I _sens(x，y，z)-min[I _sens(x，y，z)]}/

{max[I _sens(x，y，z)]-min[I _sens(x，y，z)]} (2)

In addition, in the MRI device, when carrying out parallel imagings (PI) and raw image data S by a plurality of multiple coils of use _OrigWhen the signal from multiple coil is unfolded, being synthesized, can be used to determine at the sensitivity of the multiple coil of considering to have coil independence or the g-factor distribution g (x by the noise profile under the situation of the unfolded synthetic influence of signal from being prescribed, y z) obtains weighting function W by formula (3) expression _Snr(x, y, z).

W _snr(x，y，z)＝1/g(x，y，z) (3)

Then, at step S2, sensitivity correction unit 2 obtains raw image data S as the target of sensitivity correction from image data memory cell 9 _Orig, on the other hand, obtain the sensitivity map I that is used in sensitivity correction from sensitivity map memory element 10 _Sens, and use the sensitivity map I that is obtained _SensTo raw image data S _OrigCarry out the sensitivity correction of pick off 7, obtain sensitivity correction view data S thus _ScorThen, sensitivity correction unit 2 is the sensitivity correction view data S that obtains like this _ScorBe provided to filter cell 4 and weighting summation unit 5.

Sensitivity correction view data S _ScorCan be generated according to formula (4).

S _scor(x，y，z)＝S _orig(x，y，z)/I _sens(x，y，z) (4)

I wherein _Sens(x, y, z): the sensitivity map datum

S _Orig(x, y, z): raw image data (view data before sensitivity correction), and

S _Scor(x, y, z): the view data after the sensitivity correction

Then, at step S3, filter cell 4 uses even wave filter for by sensitivity correction view data S _ScorThe sensitivity correction data that on k space or FREBAS space, obtain of conversion, or for from the sensitivity correction unit the 2 sensitivity correction data that receive carry out Filtering Processing, generate filtered view data S thus _Scor.filOr filtered data.When generation is different from filtered view data S _Scor.filFiltered data the time, the data of filtering are converted into filtered view data S _Scor.fil

Then, filter cell 4 is the filtered view data S that obtains like this _Scor.filOffer weighting summation unit 5.As a result, weighting summation unit 5 has at least at sensitivity correction view data before the Filtering Processing and the filtered image data S after Filtering Processing _Scor.filGenerate a plurality of filtered view data S in filtering according to the different intensity in filter cell 4 _Scor.filUnder the situation of section, with the filtered view data S of the number of types same number of filtering strength _Scor.filSection is provided for weighting summation unit 5.

Here, the explanation of simple example calculation will be provided, simultaneously assumed sensitivity image correcting data S _ScorBe made up of two components, the component that only has bad SNR is just smoothed by the filtering of filter cell 4.In fact, the component that only has bad SNR is smoothed just enough.

When H represents filter operator, filtered view data S _Scor.filCan obtain view data, wherein smoothing filter is applied to whole sensitivity correction view data S according to formula (5) _Scor

S _scor.fil(x，y，z)＝H[S _scor(x，y，z)] (5)

Then, at step S4, weighting summation unit 5 receives the weighting function W that distributes according to SNR from SNR distribution acquiring unit 3 _Snr, and use weighting function W _SnrThe synthetic single or multiple filtered view data S that receives from filter cell 4 _Scor.filSection, and by weighting summation from the sensitivity correction unit 2 sensitivity correction view data S that receive, before Filtering Processing _Scor, generate heterogeneity filtered image data S thus _{Scor.nonuni.fil}

Use weighting function W _SnrWeighting shown in formula (6-1) and formula (6-2), be performed.That is, weighting is so used, and makes only the component with bad SNR to be carried out filtering.As the result of this weighting, sensitivity correction view data S _ScorBasically be divided into component S with good SNR _Scor.h(x, y is z) with the component S with filtering of bad SNR _Scor.l.fil(x, y, z).

S _scor.h(x，y，z)＝W _snr(x，y，z)*S _scor(x，y，z) (6-1)

S _scor.l.fil(x，y，z)＝{1-W _snr(x，y，z)}*S _scor.fil(x，y，z) (6-2)

Subsequently, the component S that has good SNR _Scor.h(x, y is z) with the filtered component S with bad SNR _Scor.l.fil(z) these two components are synthetic mutually as shown in Equation (7), obtain being subjected to view data (the filtered view data of the heterogeneity) S that non-homogeneous SNR proofreaies and correct Filtering Processing thus for x, y _{Scor.nonuni.fil}(x, y is z) as the correction of a final proof image.

S _{scor.nonuni.fil}(x，y，z)＝?S _scor.h(x，y，z)+S _scor.l.fil(?x，y，z) (7)

The filtered view data S of the heterogeneity that obtains like this _{Scor.nonuni.fil}Be written to the image data memory cell 9 of image diagnosing system 6.After this, display unit 11 shows the filtered view data S of heterogeneity that reads from image data memory cell 9 on display _{Scor.nonuni.fil}As a result, the user can confirm the filtered view data S of heterogeneity _{Scor.nonuni.fil}Be subjected to sensitivity correction, made SNR distribute and become uniformly.

Only should be pointed out that filtered image data S by even filter filtering with varying strength _Scor.filCan be set to the target of weighting summation.In other words, as the component S with good SNR of the target of weighting summation _Scor.h(x, y, filtering strength z) can be set to be different from 0 intensity.In this case, sensitivity correction view data S _ScorBe not provided to weighting summation unit 5 from sensitivity correction unit 2.

In other words, data correction apparatus 1 with said structure adapts to the estimation that the spatial distribution of using relevant sensitivity such as the such pick off 7 of coil or time distributes or the information of measurement, by mainly being level and smooth even filtering for the data after sensitivity correction with varying strength such as view data according to the heteropical degree of sensitivity, and generate a plurality of data segments, and carry out mutual weighting summation according to the SNR distributed intelligence for the data segment of such generation, so that SNR and level and smooth intensity have opposite dependency.That is,, be not to use the common wave filter that under SNR distributes constant situation, is used for data for filtering, carry out such addition for data with little SNR, it is big to make that weight with strong level and smooth data becomes, and on the contrary, becomes little with the weight of weak level and smooth data.Should be pointed out that the data of wherein not carrying out Filtering Processing can be looked at as the data of wherein carrying out the Filtering Processing with zero intensity.

For this reason, according to data correction apparatus 1, the correction that the space non-uniform sensitivity of pick off 7 distributes is performed by simple processing, and the spatially uniform that keeps SNR to distribute simultaneously thus, might obtain even view data.

Up to now, as mentioned above, when carrying out sensitivity correction for data, a problem is arranged: i.e. SNR distribution becomes heterogeneous.In order to overcome this problem, can consider that the wave filter that has the spatial weighting of variation by use carries out smoothly for data.In this case, in order to determine the weighter factor of wave filter, must obtain noise profile or obtain noise profile in advance by from original image, extracting low frequency component by the prescan that separates.For example, in the MRI device, can use the sensitivity profile of the coil of measuring by prescan and according to the independence noise profile that determine, that be called as g-factor of multi-thread circle.Then, can consider to change smoothly the method for the weighter factor of wave filter according to noise profile function in the real space.

Yet the method according to the weighter factor that changes the real space median filter promptly, according to the core of noise profile function, has a problem in the time of on demand: promptly processing and filter construction are complicated.Especially, when the support size of wave filter was very big, the processing time occurring increased, and is used for the thickening of processing of image border.In addition, according to this method, non-homogeneous SNR can be corrected, but is difficult to obtain such as optimum filter weight distribution or the level and smooth such parameter of intensity, and the SNR that is difficult to make the filter weight distribution follow best for each data variation distributes.

On the contrary, data correction apparatus 1 shown in Figure 1 adopts a kind of bearing calibration, wherein data are corrected with simple processing, have the target that data that even SNR distributes are set to weighting summation such as the filtering of using common even wave filter simultaneously, rather than change the core of wave filter according to distribute situation ground one by one of SNR.Evenly wave filter is to use the wave filter of the high universalizable of same core, and the SNR that wherein needn't relate to space or time distributes.

In addition, even when the even wave filter with same level and smooth intensity is used to some data as the filtering target, the SNR of a plurality of data segments that carry out filtering with different level and smooth intensity after according to Filtering Processing distributes and is weighted addition respectively, so, when considering each data that obtains by addition, level and smooth intensity has spatial distribution according to SNR and the time distributes.So, be equivalent to the filtering that changes the filtering that the non-homogeneous wave filter of core intensity carries out by the SNR that distributes according to space or time by using even wave filter, can carrying out.

For this reason, needn't use other substep wave filter to be used for pretreatment and adjusting, so might carry out filtering by the even wave filter that uses single type provides optimized image.As a result, according to data correction apparatus 1, not only filter construction is better simply relatively, and the installation of wave filter is easier relatively, and can carry out high-speed processing.In other words, according to data correction apparatus 1, can avoid increasing such problem such as the complexity and the processing time of above-mentioned filter construction.

In addition, if structure adaptive filter, the filter adaptation that obtains by the SNR sef-adapting filter of Wiener wave filter representative with by combinative structure sef-adapting filter and SNR sef-adapting filter are in the even wave filter that uses in data correction apparatus 1, then can control filter characteristic best, and absorb the variation of the SNR distribution of target data simultaneously.

Should be pointed out that wave filter is performed if data filtering is to use simple linear space constant (LSI), the deterioration of spatial resolution then after Filtering Processing, occurs, therefore may the span or non-homogeneous data of time.

In view of foregoing, especially, if the deterioration of the spatial resolution wave filter that can be minimized wherein, promptly such as the such wave filter of Wiener wave filter, or wherein image space is not divided into a plurality of sections and the real space and is retained basically, and the structure adaptive filter of considering noise profile simultaneously is used to even wave filter, distributes or data that the time distributes even then have living space for SNR wherein, also can avoid such as worsening such the problems referred to above in Filtering Processing rear space resolution.In addition, if use such as structure adaptive filter or the such wave filter of Wiener wave filter according to spatial distribution or the time distribution of SNR, then SNR can be modified.

Filtering Processing for the data that obtain in the MRI device not only can and can be carried out in the k-space in the r-space.For this reason, especially, when the multi-thread circle in passing through use MRI device is carried out as the parallel imaging of high speed imaging method, can realize reduction in processing time.For example, under the situation of the processing of SMASH (the obtaining space harmonics synchronously) type of carrying out type such such as GRAPPA (general automatic calibration local parallel obtains), carry out date processing in the k space, Filtering Processing can be carried out in the k space, therefore can carry out high speed processing.In addition, even when the processing of carrying out according to SENSE (sensitivity encoding) type, the multiple that is used to carry out FFT (fast fourier transform) is 2, so processing speed generally speaking is high.

Like this, evenly wave filter is the processing that is easy to be used for complex space, and improve aspect of performance at low SNR part SNR and be better than processing in the absolute value space, so it is being favourable aspect the installation with respect to the MRI device that wherein is difficult to carry out complicated date processing in the r space.

And, data correction apparatus 1 can on the pick off 7 for SNR wherein in time be that constant normal data are carried out sensitivity correction on the space.This is that and only the filtering data by even wave filter becomes data after sensitivity correction because the weighter factor that is subjected to each data of filtering when being smooth when sensitivity becomes constantly.For this reason, if can obtain the sensitivity profile of pick off 7 in the processing of in data correction apparatus 1, carrying out, then needn't consider about the sensitivity profile state of constant sensitivity profile whether.So, be very high in the workability aspect the wave filter installation.

Fig. 3 is the functional block diagram that shows data correction apparatus according to a second embodiment of the present invention.

In data correction apparatus 1A shown in Figure 3, comprise data division unit 12 and addition unit 13 rather than weighting summation unit 5, its structure is different from the structure of data correction apparatus shown in Figure 11.Other structure of data correction apparatus 1A does not have different with the structure and the operation of data correction apparatus 1 shown in Figure 1 with operation basically.So identical Reference numeral is affixed to the parts identical with data correction apparatus 1, and omit their explanation.

Particularly, data correction apparatus 1A also comprises data division unit 12 and addition unit 13 except sensitivity correction unit 2, SNR distribution acquiring unit 3 and filter cell 4.Then, sensitivity correction unit 2 is configured to the sensitivity correction view data is provided to data division unit 12, and SNR distribution acquiring unit 3 is configured to the distributed intelligence of relevant SNR is provided to data division unit 12.

Data division unit 12 have according to relevant with the view data that obtains from SNR distribution acquiring unit 3, about the distributed intelligence of SNR, the sensitivity correction data of obtaining from sensitivity correction unit 3 generate the function of a plurality of sensitivity correction picture content data segments, and the part of the sensitivity correction picture content data that generate like this are provided to filter cell 4 and the remainder of sensitivity correction picture content data or another part are provided to the function of addition unit 13.More particularly, data division unit 12 uses the weighting summation function that the sensitivity correction data are divided into the sensitivity correction picture content data with bigger SNR and have the sensitivity correction picture content data of less SNR on image space.Single or multiple sensitivity correction picture content data with bigger SNR are provided to addition unit 13, and on the other hand, single or multiple sensitivity correction picture content data with less SNR are provided to filter cell 4.

Filter cell 4 is configured to use even wave filter that the sensitivity correction component data that the conversion of sensitivity correction picture content data with less SNR or the sensitivity correction picture content data by having less SNR obtains is carried out filtering, generates filtered view data or filtering data thus.Then, the filtered view data that obtains of filtered view data that generates like this or the conversion by filtering data is provided to addition unit 13.

Addition unit 13 has the synthetic addition that is used for by sensitivity correction picture content data that receive from data division unit 12 and the filtered view data that receives from filter cell 4, generate the function of the filtered view data of heterogeneity of the view data be equivalent to the Filtering Processing that is subjected to using non-homogeneous wave filter basically, and the function that the filtered view data of heterogeneity that generates like this is written to the image data memory cell 9 of image diagnosing system 6.

Then, the operation of data correction apparatus 1A and the explanation of action will be provided.

Fig. 4 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus 1A shown in Figure 3 for the view data execution sensitivity correction of obtaining from image diagnosing system 6 simultaneously.Each step that on Fig. 4, comprises the symbolic representation flow chart of S and numeral.Should be pointed out that identical Reference numeral is affixed to each step of the step that is equivalent to flow chart shown in Figure 2, and the detailed description of step of equal value is omitted.

At first, at step S1, SNR distribution acquiring unit 3 is according to the sensitivity map I that obtains from sensitivity map memory element 10 _SensCalculate weighting function W _Snr, and the weighting function W that obtains like this _SnrBe provided to data division unit 12.This weighting function W _Snr(x, y z) can obtain according to aforesaid the whole bag of tricks.Then, such weighting function W that obtains _Snr(x, y z) are used in data division unit 12 according to SNR distribution dividing data.

Then, at step S2, sensitivity correction unit 2 uses the sensitivity map I that obtains from sensitivity map memory element 10 _Sens, for the raw image data S that obtains from image data acquisition unit 9 _OrigCarry out sensitivity correction, obtain sensitivity correction view data S thus _ScorThen, sensitivity correction unit 2 is the sensitivity correction view data S that obtains like this _ScorBe provided to data division unit 12.

Sensitivity correction view data S _Scor(x, y z) can be generated according to formula (8).

S _scor(x，y，z)＝S _orig(x，y，z)/I _sens(?x，y，z?) (8)

Then, at step S10, data division unit 12 is used the weighting function W that obtains from SNR distribution acquiring unit 3 _Snr, the 2 sensitivity correction view data that receive are divided into a plurality of sensitivity correction picture content data segments from the sensitivity correction unit according to the big wisp of SNR.Then, data division unit 12 is the sensitivity correction picture content data S with bigger SNR _Scor.hBe provided to addition unit 13, on the other hand, sensitivity correction picture content data S with less SNR _Scor.lBe provided to filter cell 4.

Sensitivity correction view data S _Scor(x, y is z) by using weighting function W _Snr(x, y, the component of window z) divide and can carry out according to formula (9-1) and formula (9-1).According to formula (9-1) and formula (9-1), sensitivity correction view data S _Scor(x, y z) are divided into two data, sensitivity correction picture content data S _Scor.h(x, y is z) with sensitivity correction picture content data S _Scor.l(x, y, z).

S _scor.h(x，y，z)＝S _scor(x，y，z)*W _snr(x，y，z) (9-1)

S _scor.l(x，y，z)＝?S _scor(x，y，z)*{1-W _snr(x，y，z)} (9-2)

Should be pointed out that S _Scor.h(x, y z) are the sensitivity correction picture content data with good SNR, and S _Scor.l(x, y z) are the sensitivity correction picture content data with bad SNR.

Then, at step S3, filter cell 4 uses even wave filter to be divided into the sensitivity correction picture content data S with little SNR that unit 12 receives from data _Scor.lOr by sensitivity correction picture content data S _Scor.lThe sensitivity component data that obtains of conversion carry out Filtering Processing, generate the picture content data S of Filtering Processing thus _Scor.l.filOr the component data of Filtering Processing.

That is, for example, the smoothing filter shown in the formula (10) only is applied to the sensitivity correction picture content data S with bad SNR _Scor.l(x, y z), generate the picture content data S of Filtering Processing thus _Scor.l.fil(x, y, z).

S _scor.l.fil(x，y，z)＝H[S _scor.l(x，y，z)] (10)

Wherein H represents filter operator.

Then, filter cell 4 is the picture content data S of the Filtering Processing that obtains _Scor.l.filOr the picture content data S of the Filtering Processing that obtains of the conversion by the Filtering Processing component data _Scor.l.filBe provided to addition unit 13.

Then, at step S11, addition unit 13 is the sensitivity correction picture content data S with big SNR that receives from data division unit 12 _Scor.hPicture content data S with the Filtering Processing that receives from filter cell 4 _Scor.l.filAddition is used to synthesize, and generates the filtered view data S of heterogeneity thus _{Scor.nonuni.fil}

This synthetic processing can be carried out according to formula (11).

S _{scor.nonuni.fil}(x，y，z)＝S _scor.h(x，y，z)+S _scor.l.fil(x，y，z) (11)

That is the sensitivity correction picture content data S that, has good SNR _Scor.h(x, y, z) and be subjected to sensitivity correction picture content data S filtering, that have bad SNR _Scor.l(x, y z) are synthesized mutually, calculate the filtered view data S of heterogeneity thus _{Scor.nonuni.fil}(x, y is z) as the correction of a final proof image.

Then, the filtered view data S of heterogeneity _{Scor.nonuni.fil}Be written to the image data memory cell 9 of image diagnosing system 6, and on the display of display unit 11, show.

In other words, above-mentioned data correction apparatus 1A is being weighted division for the view data after the sensitivity correction according to the size of SNR on the image space, image data section with little SNR carries out filtering by even wave filter with different intensity, synthetic thus divided image data.Be used to filtering if having the core of support size enough little in the real space according to even wave filter, even then carry out at first that Filtering Processing and weighting divide each the time, comprising the processing of Filtering Processing and dividing for the weighting of data is almost approximately equivalent.So,, can obtain identical effect with data correction apparatus shown in Figure 11 according to data correction apparatus 1A.

Fig. 5 is the functional block diagram that shows the data correction apparatus of a third embodiment in accordance with the invention.

In data correction apparatus 1B shown in Figure 5, the order of processing is different with the order of data correction apparatus shown in Figure 11.Other structure of data correction apparatus 1B and operation and data correction apparatus 1 shown in Figure 1 do not have difference basically.So identical Reference numeral is affixed to the parts identical with data correction apparatus 1, and omit their explanation.

Data correction apparatus 1B is equipped with sensitivity correction unit 2, SNR distribution acquiring unit 3, filter cell 4 and weighting summation unit 5.

Filter cell 4 has from image data memory cell 9 and obtains function as the raw image data of the target of sensitivity correction, carry out the function that Filtering Processing generates original data of the raw image data of Filtering Processing or Filtering Processing by original data of using even wave filter that raw image data or the conversion by raw image data are obtained, and the raw image data of the Filtering Processing that the conversion of the raw image data of the Filtering Processing that generates like this or the original data by Filtering Processing is obtained is provided to the function of weighting summation unit 5.

Weighting summation unit 5 has the raw image data of the Filtering Processing that receives for the raw image data that obtains from image data memory cell 9 with from filter cell 4 and carries out weighting summation according to the SNR distributed intelligence that receives from SNR distribution acquiring unit 3, generating the function of non-homogeneous filtered raw image data, and the function that the filtered raw image data of heterogeneity that generates like this is provided to sensitivity correction unit 2.

Sensitivity correction unit 2 has from sensitivity map memory element 10 and obtains the sensitivity map that will be used to sensitivity correction, and use the sensitivity map obtain, and has the function that the filtered view data of the heterogeneity that obtains like this is written to the image data memory cell 9 of image diagnosing system 6 to the 5 non-homogeneous filtered raw image datas that receive carry out sensitivity correction and generate the function of the filtered view data of heterogeneity from the weighting summation unit.

Then, the operation of data correction apparatus 1B and the explanation of action will be provided.

Fig. 6 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus 1B shown in Figure 5 for the view data execution sensitivity correction of obtaining from image diagnosing system 6 simultaneously.Each step that on Fig. 6, comprises the symbolic representation flow chart of S and numeral.On Fig. 6, identical Reference numeral will be equivalent to each step of the step of flow chart shown in Figure 2, and will roughly provide their explanation.So the order of Reference numeral and processing is inconsistent.

At first, at step S1, SNR distribution acquiring unit 3 is according to the sensitivity map I that obtains from sensitivity map memory element 10 _SensCalculate weighting function W _Snr, and the weighting function W that obtains like this _SnrBe provided to weighting summation unit 5.

Then, at step S3, filter cell 4 obtains raw image data S as the target of sensitivity correction from image data acquisition unit 9 _Orig, and by using even wave filter for the raw image data S that obtains _OrigOr by raw image data S _OrigThe original data that obtain of conversion carry out Filtering Processing, generate the raw image data S of filtering thus _Orig.filOr filtered initial data.

That is, for example, as shown in Equation (12), smoothing filter is applied to whole raw image data S _Orig(x, y z), generate filtered raw image data S thus _Orig.fil(x, y, z).

S _orig.fil(x，y，z)＝H[S _orig(x，y，z)] (1?2)

Wherein H represents filter operator.

Then, filter cell 4 is filtered raw image data S _Orig.filOr the filtered raw image data S that obtains of the conversion by filtered initial data _Orig.filBe provided to weighting summation unit 5.

Then, at step S4, weighting summation unit 5 uses the weighting function W that receives from SNR distribution acquiring unit 3 _SnrTo the raw image data S that obtains from image data memory cell 9 _OrigWith the filtered raw image data S that receives from filter cell 4 _Orig.filCarry out weighting summation, be used to synthesize, generate non-homogeneous filtered raw image data S thus _{Orig.nonuni.fil}

That is, at first, as formula (13-1) and formula (13-2) expression, weighting function W _Snr(x, y z) are used to from raw image data S _Orig(x, y z) generate the S with good SNR _Orig.h(x, y, z), and by using filtered raw image data S _Orig.fil(z) generation has the component S of the filtering of bad SNR for x, y _Orig.l.fil(x, y, z).That is, weighter factor is applied to filtered raw image data S _Orig.fil(x, y, z), its filtered component S that has bad SNR with generation _Orig.l.fil(x, y, z).

S _orig.h(x，y，z)＝W _snr(x，y，z)*S _orig(x，y，z) (13-1)

S _orig.l.fil(x，y，z?)＝(1-W _snr(x，y，z)}*S _orig.fil(x，y，z) (13-2)

Then, as formula (14) expression, two component S _Orig.h(x, y, z) and S _Orig.l.fil(x, y z) are synthesized, and obtain non-homogeneous filtered raw image data S thus _{Orig.nonuni.fil}(x, y is z) as the correcting image about SNR.

S _{orig.nonuni.fil}(x，y，z)＝?S _orig.h(x，y，z)+S _orig.l.fil(x，y，z) (14)

Like this, component with good SNR and the component with filtering of bad SNR are weighted, are used to synthesize, generate non-homogeneous filtered raw image data S thus _{Orig.nonuni.fil}Then, weighting summation unit 5 is the non-homogeneous filtered view data S that generates like this _{Orig.nonuni.fil}Be provided to sensitivity correction unit 2.

Then, at step S1, sensitivity correction unit 2 is by using the sensitivity map I that obtains from sensitivity map memory element 10 _SensFor the 5 non-homogeneous filtered view data S that receive from the weighting summation unit _{Orig.nonuni.fil}Carry out sensitivity correction, obtain the filtered view data S of heterogeneity thus _{Scor.nonuni.fil}

As formula (15) expression, by using sensitivity map I _Sens(x, y z) carry out sensitivity correction, and be subjected to non-homogeneous SNR and proofread and correct image Filtering Processing, after sensitivity correction, that is, and the view data S of heterogeneity filtering _{Scor.nonuni.fil}According to non-homogeneous filtered raw image data S _{Orig.nonuni.fil}(x, y z) calculate.

S _{scor.nonuni.fil}(x，y，z)＝S _{orig.nonuni.fil}(x，y，z)/I _sens(x，y，z) (15)

Then, sensitivity correction unit 2 is the filtered view data S of heterogeneity _{Scor.nonuni.fil}Be written to the image data memory cell 9 of image diagnosing system 6.After this, display unit 11 shows the view data S of the heterogeneity filtering of reading from image data memory cell 9 on display _{Scor.nonuni.fil}

In other words, above-mentioned data correction apparatus 1B adapts to after view data is carried out Filtering Processing and weighting summation with different intensity and carries out sensitivity correction.Like this, even before proofreading and correct view data is being carried out Filtering Processing and weighting summation, and when carrying out sensitivity correction at last, SNR does not change yet.So,, can obtain the effect identical with data correction apparatus shown in Figure 11 according to data correction apparatus 1B.

Fig. 7 is the functional block diagram that shows the data correction apparatus of a fourth embodiment in accordance with the invention.

In data correction apparatus 1C shown in Figure 7, the order of processing is different with the order of data correction apparatus 1A shown in Figure 3.Other structure of data correction apparatus 1C and operation and data correction apparatus 1A shown in Figure 3 do not have difference basically.So identical Reference numeral is affixed to the parts identical with data correction apparatus 1A, and omit their explanation.

Especially, data correction apparatus 1C is equipped with sensitivity correction unit 2, SNR distribution acquiring unit 3, filter cell 4, data division unit 12 and addition unit 13.

Data division unit 12 has based on the distributed intelligence about SNR of obtaining from SNR distribution acquiring unit 3, generates the function of a plurality of original image component data sections by a plurality of original image component datas that obtain from image data memory cell 9 according to the size of SNR, and the original image component data with little SNR is provided to filter cell 4 and the original image component data with big SNR is provided to the function of addition unit 13.

Then, filter cell 4 is configured to carry out Filtering Processing by the original component data of using even wave filter and necessary transfer pair original image component data or the conversion by the original image component data to obtain and generates filtered original image component data, and addition unit 13 is configured to handle and generate non-homogeneous filtered raw image data by filtered picture content data and original image component data with big SNR being carried out addition.And sensitivity correction unit 2 is configured to the image data memory cell 9 that is written to image diagnosing system 6 by the filtered view data of heterogeneity that non-homogeneous filtered raw image data execution sensitivity correction is obtained.

Then, the operation of data correction apparatus 1C and the explanation of action will be provided.

Fig. 8 shows the flow chart that is kept the process of SNR distributing homogeneity by data correction apparatus 1C shown in Figure 7 for the view data execution sensitivity correction of obtaining from image diagnosing system 6 simultaneously.Each step that on Fig. 8, comprises the symbolic representation flow chart of S and numeral.On Fig. 8, identical Reference numeral is affixed to each step of the step that is equivalent to flow chart shown in Figure 4, and will roughly provide their explanation.So the order of Reference numeral and processing is inconsistent.

At first, at step S1, SNR distribution acquiring unit 3 is according to the sensitivity map I that obtains from sensitivity map memory element 10 _SensCalculate weighting function W _Snr, and the weighting function W that obtains like this _SnrBe provided to data division unit 12.

Then, at step S10, data division unit 12 is used the weighting function W that obtains from SNR distribution acquiring unit 3 _Snr, the raw image data S that obtains from image data memory cell 9 according to the big wisp of SNR _OrigBe divided into a plurality of original image component data sections.

By using weighting function W _Snr(x, y, the raw image data S of window z) _Orig(x, y, component z) divide and carry out shown in formula (16-1) and formula (16-2).Then, raw image data S _Orig(x, y z) are divided into two components of original image component data, promptly have the component S of big SNR _Orig.h(x, y is z) with the component S with little SNR _Orig.l(x, y, z).

S _orig.h(x，y，z)＝S _orig(x，y，z)*W _snr(x，y，z) (16-1)

S _orig.l(x，y，z)＝S _orig(x，y，z)*{1-W _snr(x，y，z)) (16-2)

Then, data division unit 12 is the original image component data S as the component with big SNR _Orig.hBe provided to addition unit 13, on the other hand, original image component data S as component with little SNR _Orig.lBe provided to filter cell 4.

Then, at step S3, filter cell 4 is by using even wave filter for from data division unit 12 original image component data S that receive, that have little SNR _Orig.lOr by original image component data S _Orig.lThe raw image data that obtains of conversion carry out Filtering Processing, generate filtered original image component data S thus _Orig.l.filOr filtered original component data.

That is, for example, as formula (17) expression, smoothing filter only is applied to the original image component data S with bad SNR _Orig.l(x, y z), obtain filtered original image component data S thus _Orig.l.fil

S _orig.l.fil(x，y，z)＝H[S _orig.l(x，y，z)] (17)

Wherein H represents filter operator.

Then, filter cell 4 is the filtered original image component data S that obtains like this _Orig.l.filOr the filtered original image component data S that obtains of the conversion by filtered original image component data _Orig.l.filBe provided to addition unit 13.

Then, at step S11, addition unit 13 is the original image component data S with big SNR that receives from data division unit 12 _Orig.hOriginal image component data S with the filtering that receives from filter cell 4 _Orig.l.filAddition is used to synthesize, and generates non-homogeneous filtered raw image data S thus _{Orig.nonuni.fil}

Has the original image component data S of big SNR _Orig.h(x, y is z) with filtered original image component data S _Orig.l.fil(z) building-up process between is carried out according to formula (18) for x, y.Then, component with good SNR and the filtered component with bad SNR are weighted synthetic,, obtain non-homogeneous filtered raw image data S by this synthetic processing _{Orig.nonuni.fil}(x, y, z), as correcting image about SNR.

S _{orig.nonuni.fil}(x，y，z)＝S _orig.h(x，y，z)+S _orig.l.fil(x，y，z) (18)

Then, addition unit 13 is non-homogeneous filtered raw image data S _{Orig.nonuni.fil}Be provided to sensitivity correction unit 2.

Then, at step S2, sensitivity correction unit 2 uses the sensitivity map I that obtains from sensitivity map memory element 10 _SensFor the non-homogeneous filtered raw image data S that receives from addition unit 13 _{Orig.nonuni.fil}Carry out sensitivity correction, obtain the filtered view data S of heterogeneity thus _{Scor.nonuni.fil}

This sensitivity correction is carried out according to formula (19), as being subjected to the filtered view data S of heterogeneity that sensitivity correction and non-homogeneous SNR proofread and correct the image of Filtering Processing _{Scor.nonuni.fil}(x, y is z) from non-homogeneous filtered raw image data S _{Orig.nonuni.fil}(x, y, z), by using sensitivity map I _Sens(x, y z) calculate.

S _{scor.nonuni.fil}(x，y，z)＝S _{orig.nonuni.fil}(?x，y，z)/I _sens(x，y，z) (19)

Then, the view data S of the heterogeneity filtering that obtains like this _{Scor.nonuni.fil}Be written to the image data memory cell 9 of image diagnosing system 6, and on the display of display unit 11, show.

In other words, above-mentioned data correction apparatus 1C adapts to and divides the back in weighting and carry out sensitivity correction for view data, carries out Filtering Processing and addition is synthetic with different intensity.As mentioned above, Filtering Processing and weighting are handled and when carrying out sensitivity correction at last, SNR does not change yet even carry out for view data before correction.So,, can obtain the effect identical with data correction apparatus 1A shown in Figure 3 according to data correction apparatus 1C.

In in the above-described embodiment data correction apparatus 1,1A, 1B and 1C, can change arbitrarily as sensitivity correction, the order of handling these three processing with the Filtering Processing and the weighting of even wave filter.

Should be pointed out that simplification angle, in some cases, preferably can before sensitivity correction, carry out with the Filtering Processing of even wave filter from Filtering Processing.In view of above content, provided the explanation of the example calculation of filter function under the situation that WIENER wave filter that data in the FREBAS space are set to target is used as even wave filter.

(x, y z) are the real space (x, y, z) space that almost completely keeps wherein in the three-dimensional FREBAS space that is generated.So when using the Wiener wave filter, the power of noise is not set to constant, and the power P n of noise is treated to FREBAS space (x, y, function z).Filtering with the Wiener wave filter can be carried out before or after sensitivity correction, but aspect Filtering Processing, the power of noise preferably is set to constant.In view of above-mentioned content, can before sensitivity correction, carry out with the filtering of Wiener wave filter, and can handle the power P n that is set to constant noise.

Promptly, when the filtering of using the Wiener wave filter was carried out before sensitivity correction, as expression in formula (20), the filter function WF of Wiener wave filter (x, y z) can be according at FREBAS space (x, y, z) (z) the power P n with noise is determined the signal intensity Ps of view data in for x, y.

WF(X，Y，Z)＝Ps(X，Y，Z)/{Ps(X，Y，Z)+Pn} (20)

On the other hand, when the filtering of using the Wiener wave filter was carried out before sensitivity correction, noise power Pn spatially changed, determine noise power Pn=Pn (x, y, z), and filter function WF (x, y z) are determined shown in formula (21).

WF(X，Y，Z)＝Ps(X，Y，Z)/{Ps(X，Y，Z)+pn(X，Y，Z)} (21)

((x, y z) are determined according to formula (22-1) and formula (22-2) the power P n of noise z) can be equivalent to the weighting function W of the inverse of sensitivity profile by use for x, y.

W(X，Y，Z)＝FR[W(X，Y，Z)] (22-1)

Pn(X，Y，Z)＝W(X，Y，Z)*?Pn′ (22-2)

FR[wherein] expression FREBAS conversion, and Pn ' is illustrated in the noise power of the end in FREBAS space (or k space).That is, (x, y z) can be according to ((the noise power Pn ' of the end in (or k space) obtains the weighting function W that FREBAS conversion z) obtains z) with in the FREBAS space for x, y for x, y by weighting function W for the power P n of the noise of formula (21).

By the way, as mentioned above, data correction apparatus 1,1A, 1B or 1C can be added to or be structured in apparatus for measuring biological information or the image diagnosing system.In view of foregoing,, be structured in the MRI device for data correction apparatus 1A shown in Figure 3 explanation and the sensitivity correction of the image that multi-thread circle obtains when being set to pick off is handled as a specific example.

Fig. 9 is the structure chart that shows MR imaging apparatus according to an embodiment of the invention.

MR imaging apparatus 20 comprises the static field magnet 21 that is used to generate magnetostatic field, is set at pad crack coil 22, gradient coil unit 23 and RF coil 24 in the columniform static field magnet 21.Static field magnet 21, pad crack coil 22, gradient coil unit 23 and RF coil 24 are structured in (not shown) in the crane boom.

MR imaging apparatus 20 also comprises control system 25.Control system 25 comprises magnetostatic field power supply 26, gradient power supply 27, pad crack coil power 28, transmitter 29, receiver 30, sequence controller 31 and computer 32.The gradient power supply 27 of control system 25 comprises X-axis gradient power supply 27x, Y-axis gradient power supply 27y and Z axial gradient power supply 27z.Computer 32 comprises input equipment 33, monitor 34, operating unit 35 and memory element 36.

Static field magnet 21 is communicated by letter with magnetostatic field power supply 26.Magnetostatic field power supply 26 provides current to static field magnet 21, to be implemented in the function that generates magnetostatic field in the imaging region.Static field magnet 21 comprises superconducting coil in many cases.Static field magnet 21 obtains electric current from magnetostatic field power supply 26, and this magnetostatic field power supply 26 is communicated by letter with static field magnet 21 when excitation.Yet in case excitation, static field magnet 21 is just kept apart with magnetostatic field power supply 26 usually.Static field magnet 21 can comprise permanent magnet, and this makes that magnetostatic field power supply 26 is unnecessary.

Static field magnet 21 has cylindrical pad crack coil 22 in inside.Pad crack coil 22 is communicated by letter with pad crack coil power 28.Pad crack coil power 28 provides current to pad crack coil 22, so that magnetostatic field becomes uniformly.

Gradient coil unit 23 comprises X-axis gradient coil unit 23x, Y-axis gradient coil unit 23y and Z axis gradient coil unit 23z.Among X-axis gradient coil unit 23x, Y-axis gradient coil unit 23y and the columniform Z axis gradient coil unit 23z each is set in the static field magnet 21.Gradient coil unit 23 also have it inner form as the bed 37 in the zone of imaging region.Bed 37 supporting object P.RF coil 24 can be provided with around bed 37 or object P, rather than is structured in the crane boom.

Gradient coil unit 23 is communicated by letter with gradient power supply 27.The X-axis gradient coil unit 23x of gradient coil unit 23, Y-axis gradient coil unit 23y communicate by letter with Z axial gradient power supply 27z with X-axis gradient power supply 27x, Y-axis gradient power supply 27y respectively with Z axis gradient coil unit 23z.

X-axis gradient power supply 27x, Y-axis gradient power supply 27y and Z axial gradient power supply 27z provide current to X-axis gradient coil unit 23x, Y-axis gradient coil unit 23y and Z axis gradient coil unit 23z respectively, so that be created in the imaging region gradient magnetic Gx, Gy and Gz along X, Y and Z direction.

RF coil 24 is communicated by letter with receiver 30 with transmitter 29.RF coil 24 has and the radiofrequency signal that provides from transmitter 29 is sent to object P and receives because the function of the NMR signal that generate, that be given to receiver of the nuclear spin in the object P that is subjected to the radiofrequency signal excitation.

Figure 10 is the figure of example that shows the detailed structure of RF coil shown in Figure 9.Figure 11 is the sectional view that shows the exemplary arrangement of WB coil 24a shown in Figure 10 and phased-array coil 24b.

RF coil 24 for example constitutes by sending RF coil 24 and receiving RF coil 24.Send RF coil 24 and use whole (WB) coil 24a, use phased-array coil 24b and receive RF coil 24.Phased-array coil 24b has a plurality of surface coils 24c.Surface coils 24c is connected to each receiving circuit 30a dividually.

Simultaneously, the surface coils 24c of phased-array coil 24b is arranged to about Z axisymmetric, in the peripheral region of the partial L of the interested specific region that for example comprises object P.And WB coil 24a is provided at the outside of phased-array coil 24b.Therefore, radiofrequency signal can send to object P by WB coil 24a, and can on a plurality of channels, be received by the surface coils 24c of WB coil 24a or phased-array coil 24b, and be provided to the receiving circuit 30a of receiver 30 from the NMR signal that comprises interested specific part L.

Yet RF coil 24 can be made of the desirable coil or the single coil that are applicable to various application.

The sequence controller 31 of control system 25 is communicated by letter with gradient power supply 27, transmitter 29 and receiver 30.Sequence controller 31 has storage to be described for making gradient power supply 27, transmitter 29 and receiver 30 drive and being created on the function of the sequence information of the required control information of gradient magnetic Gx, the Gy of X, Y and Z direction and Gz and radiofrequency signal by gradient power supply 27, transmitter 29 and receiver 30 according to the predetermined sequence of being stored.Above-mentioned control information comprises motion control information, as the intensity of the pulse current that should be applied to gradient power supply 27, application time at interval and apply sequential.

Sequence controller 31 also is configured to initial data is given to computer 32.Initial data is the complex data that the A/D of the NMR signal that detects by the detection of NMR signal with in receiver 30 is converted to.

Transmitter 29 has the function that RF coil 24 is provided radiofrequency signal according to the control information that provides from sequence controller 31.Receiver 30 has by detection changes the function that generates as the initial data of digitized complex data from RF coil 24 NMR signal that provides and the AD that carries out prearranged signal processing and detected NMR signal.Receiver 30 also has the function that the initial data that generates is given to sequence controller 31.

Computer 32 obtains various functions, and data correction apparatus 1A is stored in some program in the memory element 36 of computer 32 by execution operating unit 35 disposes.Parts corresponding to computer 32 can comprise some special circuit, rather than use some program.

Figure 12 is the functional block diagram that shows computer 32 shown in Figure 1.

Computer 32 with program is used as sequence controller control unit 40, image reconstruction unit 41, k spatial database 42, real space data base 43, scan control unit 44, sensitivity profile estimation unit 45, sensitivity map data base 46, image-display units 47 and data correction apparatus 1A.

Sequence controller control unit 40 have be used for according to from input equipment 33 or another unitary information by a predetermined sequence information is given to the function of sequence controller 31 with the driving of control sequence controller 31.And sequence controller control unit 40 also has to be used for receiving from the initial data of sequence controller 31 and initial data is aligned to the k sky that is formed on k spatial database 42 asks function in (fourier space).So the initial data that k spatial database 42 storage is generated by receiver 30 is as the k spatial data, and the k spatial data is aligned in the k space that is formed in the k spatial database 42.

Image reconstruction unit 41 has and is used for obtaining the k spatial data, carrying out prearranged signal and handle, rebuild such as the such real space data of view data and real space data are written to real space data base 43 function from k spatial database 42.Image reconstruction unit 41 is configured to carry out various processing for the k spatial data in the k space that is arranged in the k spatial database, handle the real space data that to rebuild the real space image data thus and will be used to estimate from the sensitivity of each surface coils 24c of k spatial data such as two dimension or three dimensional fourier transform.So, the real space data of real space data base's 43 storage such as view data.

Sensitivity profile estimation unit 45 has the real space data of reading the sensitivity estimation that is used for each surface coils 24c from real space data base 43, distribute so that estimate the space of each surface coils 24c and/or time sensitivity, be used for the synthetic function that also final result is written to sensitivity map data base 46 as the sensitivity map datum.Can carry out estimation based on known any means to sensitivity profile.Comprise and carry out the sensitivity prescan that is used for the sensitivity profile estimation and use the real space data that obtain like this to estimate that the method for estimation of sensitivity profile is feasible.For example, proposed according to by use each signal intensity that each surface coils 24c collects with at the sensitivity prescan time by use ratio between the corresponding signal intensity that WB coil 24a collects obtain the method for sensitivity profile, according to obtaining sensitivity profile and regulate method of contrast or the like simultaneously by each signal intensity that uses each surface coils 24c collection.

Therefore, sensitivity map data base 46 storage representations are corresponding to the sensitivity map datum of the sensitivity profile of each surface coils 24c.

Scan control unit 44 has a sequence that is used for the sensitivity prescan and is provided to sequence controller control unit 40 with the sequence that is used for the main scanning of image collection, carries out the function of sensitivity prescan and main scanning thus.

Image-display units 47 has from real space data base 43 reads view data so that be provided to display unit 34, thus the function of display image data on display unit 34.

Data correction apparatus 1A has said structure shown in Figure 3, and its explanation will be omitted.Should be pointed out that can use have Fig. 1, data correction apparatus 1,1B or the 1C of structure shown in 5 or 7.

Then, provide the operation of MR imaging apparatus 20 and the explanation of action.

Figure 13 shows by MR imaging apparatus shown in Figure 9 20 to obtain the image of object P and carry out the flow chart that keeps the process of SNR distributing homogeneity about the sensitivity correction of each surface coils 24c simultaneously for the image that obtains subsequently.Each step that on Figure 13, comprises the symbolic representation flow chart of S and numeral.

At first, obtain the sensitivity map datum of each surface coils 24c.For this reason, scan control unit 44 is provided to sequence controller control unit 40 to the sequence that is used for the sensitivity prescan, and the sequence that is used for the sensitivity prescan is output to sequence controller 31 from sequence controller control unit 40.After this, sequence controller 31 drives gradient power supply 27, transmitter 29 and receiver 30 according to the sequence that is used for the sensitivity prescan, place therein thus and form X-axis gradient power Gx, Y-axis gradient power Gy and Z axial gradient power Gz in the imaging region of object P, and generate radiofrequency signal.

Then, the NMR signal that generates by the nuclear magnetic resonance, NMR in object P is received by RF coil 24, and is provided to receiver 30.The NMR signal that receiver 30 receives from RF coil 24, and carry out the various signal processing that comprise the AD conversion, generate initial data thus as the NMR signal of numerical data.Receiver 30 is provided to sequence controller 31 to the initial data that generates like this.Sequence controller 31 is provided to sequence controller control unit 40 to initial data, and sequence controller control unit 40 is arranged on initial data in the k space that is formed in the k spatial database 42.Then, image reconstruction unit 41 is extracted the k spatial datas from k spatial database 42, and rebuilds the real space data that the sensitivity that is used for each surface coils 24c is estimated by the image reconstruction process process, so that be written to real space data base 43.

After this, sensitivity profile estimation unit 45 is read the real space data of the sensitivity estimation that is used for each surface coils 24c from real space data base 43, for example by such as being used for the such processing of synthetic low-pass filtering, space and/or the time sensitivity of estimating each surface coils 24c distribute, and final result is written to sensitivity map data base 46 as the sensitivity map datum.For the purpose of simplifying the description, when the assumed sensitivity map datum has when the one-dimensional space of directions X distributes, obtain the sensitivity map datum I of sensitivity shown in Figure 13, that represent each surface coils 24c _Sens(x), as the input data I nput1 that is added to data correction apparatus 1A.

Then, behind the sensitivity prescan, carry out the main scanning that is used for imaging.For this reason, scan control unit 44 is provided to sequence controller control unit 40 to the sequence that is used for main scanning, and the sequence that is used for main scanning is output to sequence controller 31 from sequence controller control unit 40.After this, to be similar to the flow process of sensitivity prescan, sequence controller 31 drives according to the sequence that is used for main scanning and control gradient power supply 27, transmitter 29 and receiver 30, collects the initial data that is used for imaging thus.The initial data of collecting is set in the k space that is formed in the k spatial database 42 as the k spatial data.

Then, image reconstruction unit 41 is extracted the k spatial data that is used for imaging from k spatial database 42, so that pass through image reconstruction process process reconstructed image data, and it is written to real space data base 43.This view data is to be in each surface coils 24c shown in Input1 under the influence of the sensitivity profile on the directions X, therefore must carry out sensitivity correction.Yet noise power is constant before sensitivity correction, if attempt to be set to constantly by carrying out sensitivity correction sensitivity, noise power becomes heterogeneous.In view of foregoing, not only must carry out sensitivity correction but also must carry out the correction of non-homogeneous SNR.For this reason, the raw image data S before sensitivity correction _OrigBe input to data correction apparatus 1A as Input2.

After this, at step S20, the SNR distribution acquiring unit 3 of data correction apparatus 1A is for example by using peak response I _Sens.maxFor the sensitivity map I that obtains from sensitivity map data base 46 _Sens(x) carry out normalization as Input1,, obtain weighting function W thus so that maximum becomes 1 _Snr(x).

W _snr(x)＝I _sens(x)/I _sens.max (23)

This weighting function W _Snr(x) be used as the window function that is used for the view data that obtains in the main scanning that is used for imaging being divided into two data segments according to high SNR and low SNR.For this reason, SNR distribution acquiring unit 3 is the weighting function W that obtains like this _Snr(x) be provided to data division unit 12.

Then, at step S21, sensitivity correction unit 2 is for the raw image data S that obtains from real space data base 43 _OrigWith the sensitivity map I that obtains from sensitivity map data base 46 _Sens(x) carry out sensitivity correction, obtain sensitivity correction view data S thus _Scor Sensitivity correction unit 2 is the sensitivity correction view data S that obtains like this _ScorBe provided to data division unit 12.

Then, at step S22, data division unit 12 is used the weighting function W that obtains from SNR distribution acquiring unit 3 _Snr(x) determine partition function (window function), be used for according to the height of SNR and low level the sensitivity correction view data S that obtains from sensitivity correction unit 2 _ScorBe divided into two components.That is, when the partition function that is set to Wh (x) and is used to generate the component of low SNR when the partition function of the component that is used to generate high SNR was set to Wl (x), partition function Wh (x) and Wl (x) were by using weighting function W _Snr(x) shown in formula (24-1) and formula (24-2), be determined.

Wh(x)＝W _snr(x) (24-1)

Wl(x)＝1-W _snr(x) (24-2)

Then, be when being used at the partition function Wh (x) that represents by solid line and the Wl (x) that is illustrated by the broken lines by the calculating shown in formula (25-1) and formula (25-2), sensitivity correction view data S _ScorBe divided into sensitivity correction picture content data S with high SNR level _{Scor, h}With sensitivity correction picture content data S with low SNR level _{Scor, l}Two components.

S _scor.h＝S _scor*Wh(x)＝S _scor*W _snr(x) (25-1)

S _scor.l＝S _scor*Wl(x)＝S _scor*{1-W _snr(x)} (25-2)

Then, data division unit 12 is the sensitivity correction picture content data S with high SNR level _{Scor, h}With sensitivity correction picture content data S with low SNR level _{Scor, l}Be provided to filter cell 4.

Then, at step S23, filter cell 4 reduces filtering to the normal noise that uses and is applied to the sensitivity correction picture content data S with low SNR level that obtains from data division unit 12 in the Filtering Processing to the even data in space _{Scor, l}Noise reduces wave filter can be by wherein do not utilize sensitivity map datum I on big meaning _Sens(x) the even wave filter in any space is formed.Reduce wave filter for noise, can use even wave filter, such as linear filter, Wiener wave filter, or structure optimization wave filter.Filtering Processing can be carried out in the k space.In this case, sensitivity correction picture content data S _{Scor, l}Become the k spatial data by linear transformation, be subjected to Filtering Processing then.Be transformed into real space data at filtered k spatial data.Then, filter cell 4 is the filtered picture content data S that obtains by Filtering Processing _Scor.l.filBe provided to addition unit 13.

Then, at step S24, addition unit 13 is the sensitivity correction picture content data S with big SNR that receives from data division unit 12 _{Scor, h}With the filtered picture content data S that receives from filter cell 4 _Scor.l.filAddition is used to synthesize, and generates the filtered view data S of heterogeneity thus _{Scor.nonuni.fil}The filtered view data S of this heterogeneity _{Scor.nonuni.fil}Be by generate with window function apply filtering with very strong intensity, have the picture content of low SNR and do not apply the picture content of filtering and carry out the view data that weighting summation is synthesized into subsequently have bigger weighter factor so that apply the picture content of filtering.Therefore, as a result of, this view data is equivalent to raw image data S _OrigThe view data that obtains when carrying out Filtering Processing different aspect reducing noise effects according to the space heterogeneity of SNR distribution.

Then, the filtered view data S of heterogeneity _{Scor.nonuni.fil}Be set to the output Output of data correction apparatus 1A, and be written to real space data base 43.After this, image-display units 47 is read the filtered view data S of heterogeneity from real space data base 43 _{Scor.nonuni.fil}, it will be provided to display unit 34, shows non-homogeneous filtered image thus on display unit 34.As a result, on display unit 34, show its execution sensitivity correction and the gauged image of SNR non-uniform Distribution.

Should be pointed out that as mentioned above, replace using data correction apparatus 1A, data correction apparatus 1 shown in Figure 1, data correction apparatus 1B shown in Figure 5, or data correction apparatus 1C shown in Figure 7 can be structured in the computer 32 of MR imaging apparatus 20.

Under data correction apparatus shown in Figure 11 is structured in situation in the computer 32 of MR imaging apparatus 20, generation applies the view data of filtering and it is not applied the view data of filtering it, and it is used window function to apply the view data of filtering according to degree of SNR and its image that does not apply filtering is synthesized mutually, so that the weighter factor of filtered view data is set to bigger numerical value when component has lower SNR.

In addition, when data correction apparatus 1B shown in Figure 5 or data correction apparatus 1C shown in Figure 7 can be structured in the computer 32 of MR imaging apparatus 20 and carried out timing about non-homogeneous SNR before sensitivity correction, the spatial distribution of noise is constant, therefore be steady state value by noise power is set, carry out sensitivity correction after filtering, this causes the simplification handled.

(emulation experiment)

Then, provide wherein by the data correction apparatus shown in Figure 11 of explanation carry out gauged simulation result at to(for) the abdomen images of the target that in the MRI device, obtains.

Figure 14 shows the ideal abdomen images S after the sensitivity correction of being supposed by data correction apparatus shown in Figure 11 in the emulation of image rectification _{Ideal_scor}Figure 15 shows the original image S before the sensitivity correction of being used by data correction apparatus shown in Figure 1 in the emulation of image rectification _OrigFigure 16 display sensitivity distribution IS _Sens, it is used to for the coil of abdominal part and profile thereof, shown in Figure 15 original image S _OrigCarry out sensitivity correction.Figure 17 shows by coil and original image S profile, shown in Figure 15 thereof for abdominal part _OrigThe abdomen images S that carries out sensitivity correction and obtain _{Ideal_scor}Figure 18 display noise distribution noise_scor, it is used in after sensitivity correction, carries out the emulation of image rectification by data correction apparatus shown in Figure 11.

Original image S shown in Figure 15, before sensitivity correction _OrigIt is the image that uses the 8-ch coil in fact to obtain for abdominal part.In addition, by with the sensitivity profile of the reality of coil for abdominal part shown in Figure 16 to original image S _OrigThe image that carries out sensitivity correction and obtain is abdomen images S shown in Figure 17 _{Orig_scor}Should be pointed out that in the profile of Figure 16, trunnion axis is represented normalized sensitivity profile, vertical axis is represented one-dimensional space position.In addition, in the profile at Figure 17, trunnion axis is represented abdomen images S _{Orig_scor}Signal intensity, vertical axis is represented its one-dimensional space position.

In addition, in order to obtain noise profile noise_scor shown in Figure 180, wherein standard deviation (SD) is that 1 Gaussian noise is provided to the image with sufficiently high SNR by emulation, and definite SNR=50.That is, pass through the ideal image data S of use before sensitivity correction at the noise before the sensitivity correction _IdealMaximum (S _Ideal), Gaussian noise and SNR (=50), shown in formula (26), be set up.

noise＝max(S _ideal)/SNR*(Gaussian?noise) (26)

Then, weighting function W _SnrNormalized by using on the serial section of sensitivity profile to(for) each coil of abdominal part to be set to, so that determine maximum max=1 and maximum min=0.

Under such condition,, use LSI wave filter and structure adaptive type DSA (directive construction self adaptation) wave filter to carry out image rectification emulation as even wave filter.

Figure 19 demonstration is passed through with even LSI wave filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile carries out that SNR proofreaies and correct and the image that obtains.Figure 20 shows by using the LSI wave filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile is carried out the image that is attended by the SNR nonuniformity correction of weighting summation and obtains.Figure 21 shows by using the homogeneous texture sef-adapting filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile carries out that SNR proofreaies and correct and the image that obtains.Figure 22 shows by using the homogeneous texture sef-adapting filter for the abdomen images S after sensitivity correction shown in Figure 17 _{Orig_scor}And profile is carried out the image that is attended by the SNR nonuniformity correction of weighting summation and obtains.

Trunnion axis representative on Figure 19,20, each profile diagram of 21 or 22 is at the signal intensity of Filtering Processing postabdomen image, and vertical axis is represented its one-dimensional space position.

Shown in Figure 20 and 22, by using LSI wave filter and DSA wave filter to carry out the correction of non-homogeneous SNR, obtain such image, therein SNR be high, by smoothly being weak near the solid line area surrounded, and therein SNR be low, by smoothly being strong near the zone of dotted line.The result, when with shown in Figure 19 and 21, carry out image that even timing obtains when comparing by using LSI wave filter and DSA wave filter, the image that the correction of non-homogeneous SNR by using the LSI wave filter obtains has such characteristic, be HFS around the zone be higher, and suppressed more strongly at the central area noise.In other words, by the correction of non-homogeneous SNR, can confirm that SNR is that fuzziness is suppressed in the high peripheral region therein, on the other hand, SNR is that SNR is enhanced in the low central area therein.

Figure 23 carries out under the situation of filtering operation the figure of the standard deviation of noise and the variation of RMSE by changing level and smooth intensity when being presented at by the gauged emulation of data correction apparatus shown in Figure 11 carries out image.

In Figure 23, trunnion axis representative is at the ratio " noise SD ratio " of the standard deviation " noise SD " of noise after the Filtering Processing with the standard deviation " noiseSDoriginal " of (after sensitivity correction) original noise before Filtering Processing, and the vertical axis representative is at the ratio " RMSE ratio " of the root-mean-square error after the Filtering Processing " RMSE " with root-mean-square error " RMSEoriginal " before Filtering Processing.Here, the RMSE after the Filtering Processing calculates according to ideal view data shown in Figure 14, after sensitivity correction.

In addition, in Figure 23, dotted line and the data of white square labelling representative when using the DSA wave filter to carry out even Filtering Processing, solid line and the data of black box labelling representative when using the DSA wave filter to carry out non-homogeneous Filtering Processing.Dotted line and the data of white circle representative when using the LSI wave filter to carry out even Filtering Processing, solid line and the data of black circle representative when using the LSI wave filter to carry out non-homogeneous Filtering Processing.

According to Figure 23, when using LSI and DSA wave filter to carry out even Filtering Processing, when level and smooth intensity increased, RMSE reduced gradually together with the standard deviation of noise.In case level and smooth intensity surpasses certain level, reduce the trend that increases with RMSE with regard to the standard deviation of representing noise.In contrast, when using the LSI wave filter to carry out non-homogeneous Filtering Processing, compare with the situation of wherein using the LSI wave filter to carry out even Filtering Processing, RMSE is modified, even, can confirm that RMSE is suppressed to less relatively level smoothly being strong and the standard deviation of noise when very little.

And, when using the DSA wave filter to carry out non-homogeneous Filtering Processing, to carry out non-homogeneous Filtering Processing with use LSI wave filter and compare, the minima of RMSE is bigger, even but when level and smooth intensity is very strong, can confirm that the deterioration of RMSE is very little.So, when using the DSA wave filter to carry out non-homogeneous Filtering Processing,, can confirm even be image in the low central area and level and smooth intensity when very strong with respect to SNR wherein, can be so that fuzziness be very little.

In addition, ideal image S _Ideal-scorIn the view data of reality, be unknown, therefore can not obtain RMSE.Therefore, show that the non-homogeneous Filtering Processing with the DSA wave filter is being best aspect the reliability of the selection that is used for filter strength.

(being used for determining the method for filter strength)

Then, provide the explanation of definite method of the level and smooth intensity that is used for above-mentioned even wave filter.As mentioned above, distribute, importantly determine the level and smooth intensity of even wave filter best according to SNR as the data of correction target.In view of foregoing, use description to determine best two kinds of methods of level and smooth intensity.

Carry out non-homogeneous filtering for data, ideally, make the RMSE of the ideal data that sets of signals becomes to distribute in the various piece of data minimize with non-homogeneous SNR distribution.Yet, because the signal distributions of ideal data is unknown, in normal process, also we can say, can not make that RMSE minimizes.On the other hand, the component of signal of data distributes and changes with data.But as what in the result of image rectification emulation, show, find when the data after sensitivity correction are carried out filtering by using the LSI wave filter, high fdrequency component is deteriorated into the degree that can not ignore, and when the data after sensitivity correction were carried out filtering by use such as the such structure adaptive filter of DSA wave filter, the deterioration of high fdrequency component can be minimized.

In view of foregoing, to illustrate when main utilization structure sef-adapting filter, be used for determining the noise SD of each data division wherein distribute be set to uniformly level and smooth intensity optimum condition, determine method for first of level and smooth intensity, under the hypothesis of the high universalizable wave filter that comprises the LSI wave filter in use, the RMSE that is used for determining ideal data wherein with respect to the sets of signals of each data division become to distribute to carry out minimized level and smooth intensity optimum condition, determine method for second of level and smooth intensity.

At first, provide the explanation of the method for determining for first of level and smooth intensity.

Usually, white noise distributes with uniform gain on the frequency axis direction in the k space.So, after using common LSI filter filtering, at the space product score value of k space median filter function with do not have the noise SD that measures in the part of real space signal and between them, have proportional relation.For simplicity, the one dimension LSI wave filter of consideration on the x direction of principal axis.Therefore, the filter function when the LSI wave filter is set to H (kx) and does not have the noise SD that measures in the part of real space signal be set to σ _nThe time, at the integrated value A of k space median filter function H (kx) _HCan be represented as formula (27), wherein a is set to proportionality coefficient.

$A_H={\∫}_{-K_x/2}^{K_x/2}H\left(k_x\right)\mathrm{dk}=a{\mathrm{\σ}}_n---\left(27\right)$

Wherein Kx represents bandwidth.

That is, under the situation that bandwidth Kx represents in discrete system be on each-Kx/2 is to the sampling frequency band of Kx/2.In addition, suppose as 1/2 the nyquist frequency of bandwidth Kx more much biggerly, and can ignore folding error than the peak frequency that target data has.

In formula (27), if given filter function H (kx) then can calculate the integrated value A at k space median filter function H (kx) _HIn addition, can from noise the part that before sensitivity correction, does not have real space signal or the noise SD from k space medium-high frequency part, measure noise SD σ _n

On the other hand, the minimum SNR part and the maximum S R SNR ratio SNRR partly that before filtering, have the data of non-homogeneous SNR distribution _1h, when the noise SD of minimum SNR part after sensitivity correction is set to σ _NlAnd the noise SD of maximum S R part is set to σ _NhThe time, can be by using the SNR distribution I that obtains from the sensitivity profile of pick off (coil) _SensBe represented as formula (28).

SNRR _lh＝(1/σ _nl)/(1/σ _nh)＝σ _nh/σ _nl＝{σ _n/max(I _sens)}/{σ _n/min(I _sens)}＝min(I _sens)/max(I _sens)

(28)

That is, needn't measure noise SD and obtain at the minima of SNR and the ratio between the maximum, and if before sensitivity correction, measure, then can obtain absolute noisiness.

Figure 24 be illustrated in by data correction apparatus carry out after the sensitivity correction and non-homogeneous filtering before the non-uniform Distribution of noise and the concept map of standard deviation.Figure 25 is that expression is by carrying out the distribution of the noise that homogenization obtains and the concept map of standard deviation with non-homogeneous filtering for the non-uniform Distribution of noise shown in Figure 24.

In Figure 24 and 25, each abscissa representation space position, each vertical coordinate is represented the power and the standard deviation of noise.And, on Figure 24 and 25, the distribution of every solid line display noise, the SD of every dotted line display noise.

As shown in figure 24, the noise before data are carried out non-homogeneous filtering spatially anisotropically distributes, and the high SNR part (high SNR) with little noise power and high SNR is wherein arranged and have the low SNR part (low SNR) of big noise power and low SNR.In addition, the noise σ in high SNR part (high SNR) _NhWith the noise σ in low SNR part (low SNR) _NlCan represent as illustrated in fig. 24.

The noise SD that shows among Figure 24 is set to evenly change together with SNR by non-homogeneous filtering as illustrated in fig. 25.That is, the SD of whole noise is reduced by non-homogeneous filtering, so as with at the noise SD σ in high SNR part (high SNR) before the non-homogeneous filtering _NhConsistent.As a result, generally, become the noise SD σ in the high SNR part (high SNR) that was equivalent to before non-homogeneous filtering equably at non-homogeneous filtered noise SD _Nh

Here, suppose using the LSI wave filter to the even filtered data of data of minimum SNR part (low SNR) with the even filtered data of data of maximum S R part (high SNR) are being synthesized mutually, have 1 and 0 weighter factor respectively, whole thus data result is subjected to non-homogeneous filtering.Then, be equivalent at even filtered SNR at the SNR that the data of minimum SNR part (low SNR) is carried out near non-homogeneous filtered data (center at Figure 25), and be equivalent to SNR before even filtering at the SNR that the data of maximum S R part (high SNR) is carried out near non-homogeneous filtered data (end portion at Figure 25).In other words, the SNR of the minimum SNR part (low SNR) before non-homogeneous filtering is equivalent to the SNR before even filtering, and the SNR of the part of the maximum S R before non-homogeneous filtering (high SNR) is equivalent at even filtered desirable SNR.

Here, when using LSI carry out even filtering before and after SNR between ratio be set to SNRR _Fil.lhThe time, SNR ratio SNRR _Fil.lhCan be by using the integrated value A of the filter function that is applied to minimum SNR part (low SNR) and maximum S R part (high SNR) respectively _HlAnd A _HhBe represented as formula (29).

SNRR _fil.lh＝A _hl/A _Hh (29)

So, obtain wherein the trial that level and smooth intensity optimum condition is set to the optimum condition of as shown in figure 25 " having the condition that noise SD in each data division of non-homogeneous SNR becomes the noise SD in the maximum S R part ", simplified the filter function H1 (kx) that determines to be applied to minimum SNR part (low SNR) integrated value so that the right end portion of formula (28) equate or proportional problem mutually with the right end portion of formula (29).In other words, the integrated value of filter function should be controlled such that the distribute inverse of the noise SD that is set to the part that does not have real space signal after sensitivity correction as the SNR of the data of target filtering, and the minima that distributes of SNR and ratio between the maximum become the filter function that coexists and is added to SNR wherein and becomes the integrated value of part of minima and filter function and be added to SNR wherein and become ratio proportional (situation that also comprises identical situation and proportionality coefficient multiple) between the integrated value of peaked part.

So according to formula (28) and formula (29), the integrated value of filter function H1 (kx) by formula (30) is determined.

A _Hl＝A _Hh*SNRR _lh＝A _Hh*min(I _sens)/max(I _sens) (30)

Here, if supposing the noise SD in maximum S R part (high SNR) is not changed by the LSI wave filter, the LSI wave filter that then is applied to maximum S R part (high SNR) can be looked at as and be equivalent to that to have gain be 1 wave filter, so be applied to the integrated value A of filter function of the LSI wave filter of maximum S R part (high SNR) _HhBy formula (31) are defined.

$A_{\mathrm{Hh}}={\∫}_0^{K_x}1d{k}_x=K_x---\left(31\right)$

So, when the formula of substitution as a result (30) of formula (31), obtain formula (32).

A _Hl＝SNRR _lh*K _x (32)

According to formula (32), if find the ratio and the frequency band Kx that samples between the noise SD in noise SD and the minimum SNR part (low SNR) in maximum S R part (high SNR), then can see, can determine to be applied to the partly integrated value A of the filter function H1 (kx) of (low SNR) of minimum SNR _HI

By the way, provide integrated value A in definite being used to shown in formula (32) _HIThe situation of filter function H1 (kx) under, constraints only is integrated value, therefore, the motility that is used for designing filter function H1 (kx) is very big.Should be pointed out that usually the function that filter function H1 (Kx) preferably gains and reduces with higher frequency component.In view of foregoing, for example, filter function H1 (kx) is set to the Hanning function, shown in formula (33).

H(k _x)＝0.5[1+cos(b _x*k _x/K _x)]∶|k _x|＜K _x/b _x；＝0∶other?wise (33)

Should be pointed out that bx represents to be used for to determine the parameter of the cut-off frequency of LSI wave filter, when bx=2, cut-off frequency becomes and equals to sample maximum/minimum frequency ± kx/2.

Figure 26 is that the filter function that is presented at the even wave filter in the data correction apparatus is the figure of the example under the situation of Hanning function.

In Figure 26, abscissa is represented frequency axis kx, and vertical coordinate is represented filter function H1 (kx).As shown in figure 26, the area of the part of being surrounded by filter function H1 (kx) and frequency axis kx is the integrated value AHl of the filter function H1 (kx) that determines down at formula (32).In addition, when parameter bx was conditioned, the cut-off frequency of LSI wave filter can be provided with arbitrarily in the scope of sampling maximum/minimum frequency ± kx/2.

When filter function H1 (kx) was defined shown in formula (33), the integrated value AHl of filter function H1 (kx) was represented as formula (34).

$A_{\mathrm{Hl}}={\∫}_{-K_x/b_x}^{K_x/b_x}0.5[1+\cos (b*k_x/K_x){]\mathrm{dk}}_x=2{\∫}_0^{K_x/b_x}0.5[1+\cos (b_x*k_x/K_x)]{\mathrm{dk}}_x=K_x/b_x---\left(34\right)$

So formula (35) can draw from formula (32) and formula (34).

SNRR _lh＝A _h/A _l＝K _x/b _xK _x＝1/b _x (35)

Then, when formula (35) is expressed with parameter b x, obtain formula (36).

b _x＝1/SNRR _lh＝σ _nl/σ _nh (36)

According to formula (36), can see that parameter b x can be by using the ratio SNRR between the noise SD in noise SD that provided by formula (28), in maximum S R part (high SNR) and the minimum SNR part (low SNR) _1hBe determined.

When the filter function H1 of LSI wave filter (kx) is determined by such method and non-homogeneous SNR proofreaies and correct when being undertaken by above-mentioned weighting summation, can obtain optimum data and simultaneously noise SD distribute and be set to uniformly.

Should be pointed out that when noise during (kx, ky, kz) median filter function H (kx, ky, integrated value A kz) in the k space with common distributed in three dimensions _HWith noise SD σ _nBetween relation be that wherein proportionality coefficient is set to a shown in formula (37).

$A_H={\∫}_0^{\mathrm{Kz}}{\∫}_0^{\mathrm{Ky}}{\∫}_0^{\mathrm{Kx}}H(k_x,k_y,k_z){\mathrm{dk}}_x{\mathrm{dk}}_z=a{\mathrm{\σ}}_n---\left(37\right)$

Here, if filter function H is (kx, ky, kz) be the function that is represented as the direct product type, shown in formula (38-1), then be applied to the filter function H (kx of minimum SNR part (low SNR) and maximum S R part (high SNR) respectively, ky, integrated value A kz) _HlAnd A _HhBe represented as formula (38-2) and formula (38-3) respectively.

H(k _x，k _y，k _z)＝H(k _x)H(k _y)H(k _z) (38-1)

A _Hh＝K _xK _yK _z (38-2)

A _Hl＝SNRR _lh*K _xK _yK _z (38-3)

Should be pointed out that the LSI wave filter that is applied to maximum S R part (high SNR) is assumed to be that to have gain be 1 wave filter.

According to formula (38-3), can see, under the situation of pressing the one dimension distribution at noise, filter function H (kx, ky, integrated value A kz) _HlAnd A _HhBe according to the ratio SNRR between the noise SD in noise SD that provide by formula (28), in maximum S R part (high SNR) and the minimum SNR part (low SNR) _1hObtain.

Especially, (x, y when z) the Hanning function that has parameter b x, by and a bz by use is defined, obtain formula (39) as filter function H

SNRR _lh＝A _h/A _l＝(K _x/b _xK _x)(K _y/b _yK _y)(K _z/b _zK _z)＝1/b _xb _yb _z) (39)

So the product bxbybz of Hanning function parameters bx, by and bz can obtain according to formula (39).Here, if (x, y are initial point symmetric form functions z), then can determine bx=by=bz=b by the three-dimensional filter function H of Hanning function definition.Therefore, formula (39) is represented as formula (40).

SNRR _lh＝1/b ³ (40)

So according to formula (40), (parameter b z) can be according to the ratio SNRR between the noise SD in noise SD in maximum S R part (high SNR) and the minimum SNR part (low SNR) for x, y for filter function H _1hBe determined uniquely.

More than having described at optimum condition is that the best that noise SD is set to be used under the uniform situation the level and smooth intensity of even wave filter is determined method.In addition, in order to reach with respect to the coupling of the vision optimality of view data or to obtain and the dependency of absolute SNR, can introduce a coefficient.This coefficient can be constant or variable.

For example, average SNR is set to SNRm, and SNRm is (41) expression by formula.

SNRm?＝S(DC)/sn (41)

Should be pointed out that S (DC) is illustrated near the absolute value meansigma methods of the signal of DC in the k space.That is, SNRm is set near the absolute value meansigma methods S (DC) and noise SD σ of signal the DC in the k space _nBetween ratio.

Then, introduce to use the coefficient C (SNRm) as parameter, and coefficient C (SNRm) is set to such function of SNRm as the SNRm of absolute SNR so that when SNRm more hour level and smooth intensity become bigger.And formula (32) is transformed into formula (42) by coefficient of utilization C (SNRm), and is used for the integrated value A of filter function _HlCondition can be corrected.

A _Hl＝C(SNR _m)*SNRR _lh*K _x (42)

In addition, when structure adaptive filter is used in non-homogeneous SNR timing, also depend on the signal distributions of data in the real space basically at filtered noise SD.Should be pointed out that as using the situation of LSI wave filter, then the integrated value of the filter function of structure adaptive filter can be determined if noise SD is used in the flat of signal in the real space or do not have the SD in the part of signal to be defined.In the filtering of using the LSI wave filter, in less SNR part smoothly is stronger, and spatial resolution is worsened, but in the filtering of utilization structure sef-adapting filter, can obtain the homogenization of noise profile, and keep spatial resolution simultaneously, might carry out thus and more approach Utopian correction.

In other words, above-mentioned first definite method for level and smooth intensity is to determine filter function by use as the noise of the denominator of the SNR of DC component.

(application of Wiener wave filter)

Then, provide the explanation of the method for determining for second of level and smooth intensity.

Determine that for second of level and smooth intensity method is to be used for determining level and smooth intensity, so that minimize at the RMSE that uses the aforesaid high universalizable wave filter such to carry out data under the situation of filtering such as the Wiener wave filter.

When the filter function Hw of Wiener wave filter (after this being expressed as WF) can be expressed as power that power when signal is set to Ps and noise ideally and be set to Pn about the spatial function of filtering shown in formula (43).

Hw＝Ps/(Ps+Pn) (43)

Usually, the power P s of signal is for the spatial function that is applied in WF, and the power P n of noise is constant.Be applied in the total space of WF and stipulated by fourier space, the WF that is stipulated by fourier space is represented as FT-WF.Should be pointed out that the target that is applied in WF can be FREBAS space and any WF space that is divided into a plurality of resolution, the WF that is stipulated by the FREBAS space is represented as FR-WF.

Usually, carrying out timing for the data with non-homogeneous SNR, WF is not used to wherein to have filtering noise profile, after the sensitivity correction.Yet, when carry out by use after sensitivity correction SNR therein become the power of the noise in the peaked part and therein SNR become the WF processing that the power of the noise in the part of minima is optimized, and the data after WF handled are when carrying out weighting summation, can think can implementation space the best SNR proofread and correct.In view of last content, WF is applied to the data after sensitivity correction.

Under the big situation to a certain degree of SNR, the filter function Hw of WF can be confirmed as ideal type, shown in formula (43) therein.In this case, when the signal distributions of ideal data was the unknown, the power P s of signal can draw from the data as the filtering target.In addition, under the little situation to a certain degree of SNR, the filter function Hw of WF can be confirmed as threshold type therein, and shown in formula (44), therein, threshold value or lower numerical value are looked at as zero.

Hw＝max[Ps-Pn，0]/ps (44)

And when determining above-mentioned filter function Hw, the power P s of signal also can be obtained according to the dependency between adjacent voxel.In addition, the power P n of noise can be corrected shown in formula (45) by using correction coefficient Ca.

Pn＝Ca*Pn (45)

In other words, determine method according to above-mentioned for second of level and smooth intensity, when carrying out filtering by WF, the power P s of the signal in each data division is used as the approximate solution of the signal distributions of ideal data, and data are minimized with respect to the RMSE of ideal data.

Then, by adopting above-mentioned second definite method for level and smooth intensity, can be so that the level and smooth intensity optimization of wave filter.This optimization function that is used for level and smooth intensity can be provided to filter cell 4.Here, provide explanation for the handling process of the filtering relevant with level and smooth intensity optimization.For example, provide the explanation of the situation of wherein in the filter cell 4 of data correction apparatus shown in Figure 11, carrying out the filtering relevant with level and smooth intensity optimization.

Figure 27 is the even flow chart of the processing stream under the situation of the optimization filtering of the level and smooth intensity of wave filter in the filter cell that is presented in the data correction apparatus shown in Figure 11.Each step that on Figure 27, comprises the symbolic representation flow chart of S and numeral.

At first, at step S30, be transformed into data in the filtering space as the data filtering target, in the real space.When view data being carried out filtering, at the view data S in the real space after the sensitivity correction with FT-WF _Scor(x, y z) are subjected to FT, view data S _Scor(x, y z) are transformed into (kx, ky, kz) the data S in the k space _Scor(kx, ky, kz), shown in formula (46-1).Simultaneously, for the processing of after this describing, the view data S before sensitivity correction in the real space _Orig(x, y z) are subjected to FT, are transformed into (kx, ky, kz) the data S in the k space _Orig(kx, ky, kz), shown in formula (46-2).

S _scor(kx，ky，kz)＝FT[S _scor(x，y，z)] (46-1)

S _orig(kx，ky，kz)＝FT[S _orig(x，y，z)] (46-2)

Should be pointed out that when view data being carried out filtering,, and use the FREBAS conversion, be used to transform to the FREBAS space without FT with FR-WF.After this, provide the explanation of wherein view data being carried out the situation of filtering with FT-WF.

Then, at step S31, be from the power P norig of the noise before sensitivity correction and the sensitivity profile I of pick off at the minima Pnl and the maximum Pnh of noise power after the sensitivity correction _Sens(x, y z) obtain.That is, the power P norig of the noise before sensitivity correction be from before sensitivity correction at k spatial data S _Orig(kz) HFS in obtains for kx, ky.Then, according to formula (47-1) and formula (47-2), obtain the minima Pnl and the maximum Pnh of noise power after sensitivity correction.

Pnl＝Pnorig/min[I _sens(x，y，z)] (47-1)

Pnh＝Pnorig/max[I _sens(x，y，z)] (47-2)

Then, at step S32, minima Pn1 and maximum Pnh according to noise power after sensitivity correction, obtain being applied to the filter function Hwh (kx that SNR wherein becomes the WF of peaked part by formula (43) or formula (44), ky, kz) and be applied to SNR wherein become minima part WF filter function Hwl (kx, ky, kz).

Then, at step S33, shown in formula (48-1) and formula (48-2), by the filter function Hwl of two types level and smooth intensity (kx, ky, kz) and Hwh (kz) Gui Ding WF is applied to the k spatial data S after sensitivity correction for kx, ky _Scor(kx, ky, kz), thus k spatial data S _Scor(kx, ky kz) are divided into two k spatial component data segment: data S _Scor.fil.l(kx, ky is kz) with data S _Scor.fil.h(kx, ky, kz).

S _scor.fil.l(kx，ky，kz)＝Hwl(kx，ky，kz)*S _scor(kx，ky，kz) (48-1)

S _scor.fil.h(kx，ky，kz)＝?Hwh(kx，ky，kz)*S _scor(kx，ky，kz) (48-2)

Then, at step S34, shown in formula (49-1) and formula (49-2), the k spatial component data S in the filtering space _Scor.fil.l(kx, ky is kz) with k spatial component data S _Scor.fil.h(kx, ky kz) are transformed into real space component data S respectively by IFT (inverse Fourier transform) _Scor.fil.l(x, y, z) and S _Scor.fil.h(x, y, z).

S _scor.fil.l(x，y，z)＝IFT[S _scor.fil.l(kx，ky，kz)] (49-1)

S _scor.fil.h(x，y，z)＝IFT[S _scor.fil.h(kx，ky，kz)] (49-2)

Then, the real space component data S that obtains like this _Scor.fil.l(x, y is z) with real space component data S _Scor.fil.h(x, y z) are provided to weighting summation unit 5 as the dateout from filter cell 4.Then, as mentioned above, (x, y z) are used to the real space component data S weighting function Wsnr of the distribution of the SNR that representative obtains in SNR distribution acquiring unit 3 _Scor.fil.l(x, y is z) with real space component data S _Scor.fil.h(x, y z) are weighted addition, generate the view data that wherein non-homogeneous SNR distributes and is corrected thus.

By processing such in filter cell 4, the non-homogeneous filtering relevant with the optimization of level and smooth intensity can be carried out under optimum condition, wherein,, handles RMSE data simultaneously so that being minimized with non-homogeneous SNR distribution by using WF to carry out filtering.Should be pointed out that to be similar to flow process shown in Figure 2 that when the maximum S R that is set at non-homogeneous filtered SNR before non-homogeneous filtering, (z) Gui Ding WF is not applied to k spatial data S for x, y by filter function Hwh _Scor(z), and only (z) Gui Ding WF is applied to k spatial data S for x, y by filter function Hwl for x, y _Scor(x, y, z).In this case, not real space component data S _Scor.fil.h(x, y z) output to weighting summation unit 5 from filter cell 4, but the view data S after sensitivity correction _Scor(z) target as weighting summation is provided to weighting summation unit 5 from correcting unit 2 for x, y.

In addition, as mentioned above, using FT-WF to carry out the deterioration of spatial resolution to a certain degree to occur under the situation of filtering.On the other hand, if use FR-WF to carry out filtering, the deterioration of spatial resolution might be suppressed to minima.

(application examples of X ray CT device)

Each data correction apparatus 1,1A, 1B and 1C can be structured in the X ray CT device.So, will be described in the example that data for projection that 1 pair of the data correction apparatus shown in Figure 1 of structure in the X ray CT device obtains by the X-ray detector that is used as pick off or X ray CT view data are carried out sensitivity correction.

Figure 28 is the structure chart that shows X ray CT device according to an embodiment of the invention.

X ray CT device shown in Figure 28 comprises crane boom part 51 and computer part 52.Crane boom part 51 comprises X-ray tube 53, high pressure maker 54, X-ray detector 55 and DAS (data-acquisition system) 56.Figure 28 shows the multitube CT device that is equipped with two X-ray tube 53A and 53B and X-ray detector 55A and 55B.Should be pointed out that and also can use the single tube CT device that is equipped with single X-ray tube 53 and X-ray detector 55.

X-ray tube 53A and 53B and X-ray detector 55A and 55B are provided at and are sandwiched in (not shown) on the rotating ring of the relative position of intermediary object P.

High pressure maker 54 is configured to tube current and tube voltage are provided to X-ray tube 53A and 53B respectively.X-ray detector 55A and 55B are configured to detect respectively from X-ray tube 53A and 53B X ray that expose and that pass through object P transmission.And the x-ray detection signal that is detected by X-ray detector 55A and 55B is provided to and is used for digitized DAS 56 respectively, is provided to computer part 52 then.

Computer part 52 with program is used as data processing unit 57, data for projection memory element 58, CT image data memory cell 59 and detector sensitivity distributed store unit 60.And data correction apparatus 1 shown in Figure 1 is structured in the computer part 52.

Data processing unit 57 has by for carrying out the function that various date processing generate data for projection and X ray CT image from the x-ray detection signal of DAS 56.The data for projection and the X ray CT image that are produced by data processing unit 57 are stored in respectively in data for projection memory element 58 and the CT image data memory cell 59.

And, the spatial sensitivity profile information of detector sensitivity distributed store unit 60 storage each X-ray detector 55A and 55B.

Then, each spatial sensitivity profile information of being configured to by using the X-ray detector 55A obtain from detector sensitivity distributed store unit 60 and 55B of the sensitivity correction unit 2 of data correction apparatus 1 is carried out sensitivity correction to the data for projection that obtains from data for projection memory element 58 or from the X ray CT view data that CT image data memory cell 59 is obtained.

And SNR distribution acquiring unit 3 is configured to estimate the distribution of the SNR that the sensitivity correction of data for projection or X ray CT view data generated together with the X ray CT view data of obtaining by the data for projection that uses the sensitivity profile information obtained from detector sensitivity distributed store unit 60 and obtain from data for projection memory element 58 or from CT image data memory cell 59.

The space S NR of data for projection distributes and can obtain in the intensity of each channel from the X-ray detector signal of each X-ray detector 55A and 55B output.The space S NR of radioscopic image data distributes and can be obtained from the CT image of the reconstruction that produces with rough matrix.

Figure 29 be illustrated in the imaging region of X ray CT device shown in Figure 28 the position and from the figure of the relation between the intensity of the x-ray detection signal of each X-ray detector 55A and 55B output.

In Figure 29, abscissa is illustrated in the position on the imaging region, and vertical coordinate is represented the intensity of x-ray detection signal.

When the SNR that for example obtains data for projection distributed, the x-ray dose corresponding to such as the part of the structure of skeleton was shown as X-ray absorption coefficient big on the projecting direction as shown in figure 29, is reduced.So,, obtain from the intensity distributions of the x-ray detection signal of X-ray detector 55A and 55B output with respect to all data for projection.Therefore, the intensity distributions of x-ray detection signal can be used as the SNR distribution.

And the SNR distribution (SNR) of X ray CT view data is equivalent to the inverse of the CT value (CT#) of the CT image of rebuilding roughly, shown in formula (50).

1/CT#∝SNR (50)

In the X ray CT device 50 with data correction apparatus 1, inherent space non-uniform sensitivity distributes and can be corrected in each X-ray detector 55A and 55B, and the spatially uniform that keeps SNR to distribute simultaneously.In addition, in the X ray CT device 50 with data correction apparatus 1, the sensitivity difference between X-ray detector 55A and 55B also can be corrected.

In other words, when not only in multi-tube X-ray CT device 50 but also have the data of collecting in the medical treatment device of a plurality of pick offs and be set to data correction apparatus 1,1A, during the correction target of 1B or 1C, each spatial sensitivity difference and the sensitivity difference between pick off that can correcting sensor.

Claims (25)

1. data correction apparatus comprises:

The sensitivity correction unit is configured to distribute by the non-uniform sensitivity that use is used to obtain the pick off of correction target data and carries out sensitivity correction for first target data that obtains according to the correction target data and produce first deal with data; And

The SNR distribution correcting unit, be configured to produce the component data section, each component data section distributes according to SNR and is subjected to respective weight, and the corresponding filtering by using second target data that obtains according to the correction target data to have different mutually intensity, so that produce second deal with data by mixing these 1 a little component data sections.

2. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit is configured to produce the component data section with depending on the weighting division that SNR distributes or using the division of being undertaken by for the corresponding filtering with different mutually intensity of second target data.

3. according to the data correction apparatus of claim 1,

Wherein said sensitivity correction unit is configured to consider to be used as the correction target data execution sensitivity correction of first target data; And

First deal with data that described SNR distribution correcting unit is configured to be considered as correction target data second target data, after sensitivity correction produces second deal with data.

4. according to the data correction apparatus of claim 1,

The correction target data that wherein said sensitivity correction unit is configured to be considered as second target data produce second deal with data; And

Described SNR distribution correcting unit is configured to be considered as second deal with data execution sensitivity correction of first target data.

5. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit be configured to by use mutually different weighter factors the second target data weighting is produced first in the middle of the component data segment, and carry out corresponding filtering with different mutually intensity for component data segment in the middle of first.

6. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit is configured to produce the second middle component data segment by carry out the corresponding filtering with different mutually intensity for second target, and is weighted addition for the second middle component data segment.

7. according to the data correction apparatus of claim 1,

Also comprise SNR distribution estimating unit, be configured to estimate that by using non-uniform sensitivity to distribute SNR distributes,

Wherein said SNR distribution correcting unit is configured to carry out weighting according to being distributed by the SNR of described SNR distribution estimating unit estimation.

8. according to the data correction apparatus of claim 1,

Also comprise SNR distribution estimating unit, be configured to estimate that by use correction target data SNR distributes,

Wherein said SNR distribution correcting unit is configured to carry out weighting according to being distributed by the SNR of described SNR distribution estimating unit estimation.

9. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit be configured to by use have can be assumed to be on the big meaning in time be that the wave filter of uniform characteristic is carried out the corresponding filtering with different mutually intensity on the space.

10. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit is configured to be confirmed as making at least one wave filter of the optimized wave filter of SNR in the space of frequency band division to carry out the corresponding filtering with different mutually intensity by the linear filter that uses its intensity variable, structure adaptive filter and its intensity.

11. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit is configured to by using its intensity to be confirmed as to make the optimized wave filter of SNR to carry out the corresponding filtering with different mutually intensity, and to handle the space be fourier space, by applying that the Fresnel conversion is divided into the space of frequency band and by divide one of space that obtains between real-time empty with wavelet transformation.

12. according to the data correction apparatus of claim 1,

Wherein said sensitivity correction unit is configured to first target data that obtains for according to a view data in a dimensional data image, two-dimensional image data, 3 d image data and the four-dimensional image data, carries out sensitivity correction about this view data as the correction target data; And

Described SNR distribution correcting unit is configured to second target data that obtains according to this view data by using, produces second deal with data.

13. according to the data correction apparatus of claim 1,

Wherein said sensitivity correction unit is configured to first target data that obtains for according to the data with time shaft obtained by one of electroencephalogram, electrocardiogram, oscillosynchroscope and supersonic diagnostic appts, be considered as the data with time shaft of correction target data, carry out sensitivity correction; And

Described SNR distribution correcting unit is configured to second target data that obtains according to the data with time shaft by using, produces second deal with data.

14. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit is configured to obtain the power of noise on the data that first after the sensitivity correction is handled by using non-uniform sensitivity to distribute, by using its intensity to be confirmed as mutually different so that make that according to the power of noise SNR is optimized, each filter function is carried out the corresponding filtering with different mutually intensity for first data of handling, and for each by the corresponding filtering with different mutually intensity produce the 3rd in the middle of the component data segment carry out weighting summation.

15. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit is configured to obtain the peak power and the minimum power of noise on the data that first after the sensitivity correction is handled by using non-uniform sensitivity to distribute, make SNR optimized by using its intensity to be confirmed as making according to maximum noise power, the filter function that is used for minimum SNR is confirmed as making with its intensity and makes SNR optimized according to maximum noise power, each filter function that is used for the filter function of maximum S R is carried out the corresponding filtering with different mutually intensity for the data of first processing, and carries out weighting summation for each by two the 4th middle component data segments that the corresponding filtering with different mutually intensity produces.

16. according to the data correction apparatus of claim 1,

Wherein said SNR distribution correcting unit comprises the filter strength determining unit, is configured to the filter function of determining to be used to have the corresponding filter filtering of different mutually intensity according to the integrated value of filter function and the condition that SNR distributes.

17. according to the data correction apparatus of claim 16,

Wherein said filter strength determining unit is configured to consider determine filter function distributing as the SNR on the filtering target data at the inverse of the SD of the noise on the part that does not have real space signal after the sensitivity correction, so that minima that SNR distributes on as the filtering target data of the target of the corresponding filtering with different mutually intensity and the ratio between the maximum and the part that is being applied to the minimum SNR of demonstration on the filtering target data, be used for minimum SNR filter function integrated value with in the part that is applied to the demonstration maximum S R on the filtering target data, the ratio that is used between the integrated value of filter function of maximum S R is directly proportional.

18. according to the data correction apparatus of claim 16,

Wherein said filter strength determining unit is configured to depend on the condition of coming the integrated value of correcting filter function as the correction coefficient of absolute SNR of the filtering target data of the target of the corresponding filtering with mutual varying strength by use.

19. a data correcting method may further comprise the steps:

The non-uniform sensitivity that is used to obtain the pick off of correction target data by use distributes and carries out sensitivity correction for first target data that obtains according to the correction target data and produce first deal with data; With

Produce the component data section, each component data section distributes according to SNR and is subjected to respective weight, and the corresponding filtering by using second target data that obtains according to the correction target data to have different mutually intensity, so that produce second deal with data by mixing these component data sections.

20. a MR imaging apparatus comprises:

Coil;

Data capture unit is configured to obtain the magnetic resonance image data of object and at least one item in the k spatial data by the described coil that is used as pick off;

The sensitivity correction unit is configured to carry out sensitivity correction by the non-uniform sensitivity distribution of using described coil for first target data that obtains according at least one item in magnetic resonance image data and the k spatial data and produces first deal with data; And

The SNR distribution correcting unit, be configured to produce into the component data segment, each component data section distributes according to SNR and is subjected to respective weight, and the corresponding filtering by using second target data that obtains according to magnetic resonance image data and at least one item in the k spatial data to have different mutually intensity, so that produce second deal with data by mixing these component data sections.

21. according to the MR imaging apparatus of claim 20,

Wherein said coil comprises surface coils, and

Described data capture unit is configured to obtain at least one in magnetic resonance image data and the k spatial data by the described surface coils that is used as described pick off.

22. according to the MR imaging apparatus of claim 20,

Wherein said coil comprises surface coils, and

The described surface coils that described data capture unit is configured to by being used as described pick off obtains in magnetic resonance image data and the k spatial data at least one by the parallel expansion imaging.

23. the MR imaging apparatus according to claim 20 also comprises:

The sensitivity profile acquiring unit estimates that according to the magnetic resonance signal that obtains from object the non-uniform sensitivity of described coil distributes;

SNR distribution estimating unit is configured to distribute according to the non-uniform sensitivity of being estimated by described sensitivity profile acquiring unit and estimates that SNR distributes,

Wherein said SNR distribution correcting unit is configured to carry out weighting according to being distributed by the SNR of described SNR distribution estimating unit estimation.

24. according to the MR imaging apparatus of claim 20,

Wherein said coil comprises surface coils,

Also comprise SNR distribution estimating unit, be configured to estimate that according to the g-factor that the independence according to described surface coils obtains SNR distributes,

Wherein said SNR distribution correcting unit is configured to carry out weighting according to being distributed by the SNR of described SNR distribution estimating unit estimation.

25. an X ray CT device comprises:

X-ray detector;

Data capture unit is configured to obtain the view data of object and at least one item in the data for projection by the described X-ray detector that is used as pick off;

The sensitivity correction unit is configured to carry out sensitivity correction by first target data that obtains according at least one item in view data and the data for projection for the uniform sensitivity profile of using described X-ray detector and produces first deal with data; With

The SNR distribution correcting unit, be configured to produce the component data section, each component data section distributes according to SNR and is subjected to respective weight, and the corresponding filtering by using second target data that obtains according to view data and at least one item in the data for projection to have different mutually intensity, so that produce second deal with data by the mixed components data segment.

Images