CN101524278A - Method for correcting image depth effect - Google Patents

Abstract

The invention discloses a method for correcting image depth effect which is used for correcting an original image which is obtained from a positron emission tomography by a filtered back projection rebuilding algorithm. In the method, probability distribution data used for image distribution rebuilding at each point within the visual range of a detector of the positron emission tomography is beforehand obtained; then the probability distribution data is processed by the filtered back projection rebuilding algorithm; the depth effect image removal is carried out on the probability distribution data in order of from an outermost ring to an innermost ring; and removed part is transformed into a corresponding exact image to compensate the original image, so as to finish the correction of the depth effect. According to the invention, original software kernel is not required to be amended at all, the correction of DOI effect can be finished just by the superaddition based on the original software, and the invention has flexible operation and also saves time.

Description

A kind of method for correcting image depth effect

Technical field

The present invention relates to the position emissron tomography instrument, relate in particular to a kind of method for correcting image depth effect that is used for the position emissron tomography instrument.

Background technology

Position emissron tomography instrument (Position Emission Tomograpny, abbreviate PET as) on being used for the medical science contrast imaging time, at first need in the patient body, inject the biological spike medicine that includes radionuclide, after this biology spike medicine is introduced in the human body, its inner nucleic participates in the vital movement of intracorporeal organ, tissue, cell, decay can take place discharge the γ photon that a pair of energy is identical, direction is opposite, PET catches this to the γ photon then, carry out image reconstruction, thereby find the intravital focus of patient according to the image of rebuilding.

In conventional P ET technology, catch the γ photon by the crystal bar in the detector of PET, the general incoming position that adopts the inner face of crystal bar as the γ photon, this is real response line (line ofresponse to the γ photon at two lines that crystal was linked to be that ray hits, abbreviate LOR as), but concrete which position on line of response but has only by image reconstruction and just can determine.As shown in Figure 1, certain radioactive source discharges the real response line of two γ photon incident crystal bars shown in solid line, and in image reconstruction, can think position at dotted line, will cause the position deviation of real response line and image reconstruction line of response like this, and finally cause error in the imaging, we are referred to as effect of depth (Depth of Interaction is called for short DOI) this deviation.In the PET imaging of annular and polygon detector, the radioactive source off-center is far away more, and the incident inclination angle of γ photon is big more, and the DOI effect that causes is obvious more, so just causes the image quality at edge, the visual field seriously to descend.

The principle of position emissron tomography instrument imaging is the data that obtain to be used for imaging earlier by the hardware components in the scanner, and then by software section wherein these data is rebuild, thereby obtains image.Therefore, correction processing method at the effect of depth (DOI effect) of position emissron tomography instrument (PET), two kinds of methods are arranged, a kind ofly directly handle at the front end hardware components, this method generally is to adopt to carry out the stratified principle of crystal bar on hardware, learn that the γ photon specifically gets to that one deck, thereby directly from hardware instruments receive different source positions have the DOI effect and the DOI effect not obvious (be the unconspicuous data definition of DOI effect accurate data here, data Hereinafter the same) are handled the data that these data obtain being used for image reconstruction then.The method of this hardware need be revised the hardware system of instrument, has increased the cost of manufacture and the complexity of instrument.Especially for the instrument of finishing use, software system and hardware system all need significantly to revise, and this method is very inapplicable.

Another kind is to carry out the DOI effect calibration at software section to handle, collect the data that are used for imaging at hardware components after, when software section is rebuild, the DOI effect is proofreaied and correct processing.As shown in Figure 2, the correction handling process that software section carries out the DOI effect is: the data of not handling through overcorrect that hardware components is collected, be converted to the data format that is suitable for image reconstruction earlier, such as sinusoidal rectangular histogram sinogram form (abbreviating the sino data as), then the data after the conversion are proofreaied and correct processing, data after the algorithm for reconstructing that adopts software section is again handled described correction are rebuild the image after obtaining proofreading and correct.Wherein Chang Yong algorithm for reconstructing is filtered back projection's (filtered back projection is called for short FBP) algorithm for reconstructing.Describedly data after the conversion are proofreaied and correct to handle be exactly to carry out data to integrate, perhaps these data are carried out the convolution correction, perhaps utilize method of maximum likelihood and radioactive source DOI to distribute data refinement and redistribution.This software approach need be proofreaied and correct processing to being input to algorithm for reconstructing data before, and the data of the detector configuration of different PET are proofreaied and correct also difference of processing method, therefore need respectively the kernel of the software section of PET is revised at different PET detector configurations, this method operation realizes dumb, computational speed is slow, be unfavorable for practical application, and at or image value not obvious 0 zone to those DOI effects, also to lose time very much and resource through complicated trimming process.

Summary of the invention

In order to address the above problem, the purpose of this invention is to provide a kind ofly be convenient to realize, with strong points and bearing calibration that do not revise the software section kernel, this method is for the PET instrument of producing before those that does not have the DOI calibration function, as long as do the interpolation of a performance at software section, just can finish correction to the DOI effect, and this method is only to interested, the tangible regional correction of DOI effect, and operation realizes flexibly.

For this reason, the invention provides a kind of method for correcting image depth effect that is used for the position emissron tomography instrument, a kind of method for correcting image depth effect, be used to proofread and correct the original image that the position emissron tomography instrument obtains by filtered back projection's algorithm for reconstructing, it is characterized in that, said method comprising the steps of:

S1: obtain the interior every bit position of described position emissron tomography instrument vision detector scope and be used for the Probability Distribution data that imaging distributes and rebuilds;

S2: utilize described filtered back projection algorithm for reconstructing to handle described Probability Distribution data, according to original image being carried out the removal of effect of depth image to the order of innermost ring from outer shroud, and be a fractional conversion of removing that corresponding exact image compensates original image, finish the correction of effect of depth.

Wherein said step S2 may further comprise the steps further:

S21: for the selected ring of present image, whether the image that detects the arbitrary position of described selected ring is that the radioactive source imaging that exists of position forms thus, is then, changes the S22 step over to, otherwise continues to detect the next position image of described selected ring;

S22: the Probability Distribution data of extracting described position, described Probability Distribution data comprise internal layer data sino_0 and outer layer data sino_doi, while is according to the proportionate relationship of described position C=(∑ sino_doi)/(∑ sino_0), acquisition is used to compensate this selected ring produces effect of depth to next ring compensation distributed data C*A*sino_0, wherein A is the pixel value of described position, change the S21 step then over to, up to the compensation distributed data that has obtained all positions of described selected ring;

S23: the sino_doi of each position of described selected ring added up obtains sino_d, and the compensation distributed data of described each position added up obtains sino_a;

S24: the sino_d data of handling described selected ring by filtered back projection's algorithm for reconstructing obtain image image_d, and handle described sino_a data and obtain image image_a, having proofreaied and correct this image of selecting after encircling is described present image value subtracted image image_d and adds epigraph image_a, change the S21 step then over to and proofread and correct next ring image of processing, described present image is the image behind the intact last ring of described correction, up to handling all original images.

Wherein said original image is the part in the entire image that obtains of described position emissron tomography instrument, and described original image is an image value greater than zero image, is zero to other image section assignment of not proofreading and correct processing in the entire image.

9. wherein the detection method in described step S21 is: the data sino_0 that adopts filtered back projection's algorithm for reconstructing to handle described position obtains the image distribution image_0 of unit strength, if the image distribution of original image in described position comprises the image distribution of A*image_0, think that then the image of position described in the described original image is that intensity by described position is that the radioactive source of A forms.

The wherein said sinusoidal histogrammic data format of Probability Distribution The data that is used for imaging distribution reconstruction.

The every bit position is used for the Probability Distribution data that imaging distributes and rebuilds in the wherein said described detector regimes of obtaining in advance, comes analog prober to obtain by adopting GATE software.

Two-layer method was obtained described Probability Distribution data inside and outside wherein said GATE software simulation crystal divided.

As shown from the above technical solution, the present invention is by algorithm for reconstructing that directly calls the filtered back projection among original PET and the effect of depth of handling original image in conjunction with the detector data that simulation in advance obtains, to reach the purpose of proofreading and correct original image, have following beneficial effect:

1, this method is directly called original FBP algorithm for reconstructing among the PET, do not need original software section is made any modification, like this, for the instrument of producing before those that does not have the DOI calibration function, need only the interpolation of doing a performance at software section, just can finish correction the DOI effect;

2, as long as to interested, the tangible regional correction of DOI effect gets final product, and flexible operation also saves time.

3, the image resolution ratio after the correction is very high, has eliminated the DOI effect.

With reference to the accompanying drawing description of a preferred embodiment, above-mentioned and other purpose of the present invention, feature and advantage will be more obvious by following.

Description of drawings

Fig. 1 produces the principle schematic of DOI effect for the position emissron tomography instrument;

Fig. 2 is the method sketch map that software approach is handled the DOI effect in the prior art;

Fig. 3 is the method sketch map of difference the present invention and prior art;

Fig. 4 is the sketch map of the annular detector configuration of one embodiment of the invention;

Fig. 5 is an annular detector reconstructed image sketch map shown in Figure 4;

Fig. 6 carries out gauged principle schematic for topography of the present invention.

The specific embodiment

To describe specific embodiments of the invention in detail below.Should be noted that the embodiments described herein only is used to illustrate, be not limited to the present invention.

Main design of the present invention is with the image distribution probability data of simulating diverse location in the PET vision detector that obtains, is applied to carry out the correction of DOI effect on the original image of PET imaging.

Before describing the present invention in detail, difference with the prior art of the present invention is described, once earlier so that understand the present invention better.Comparison diagram 2 and Fig. 3, as can be seen, the present invention directly proofreaies and correct processing on the original image that FBP obtains, do not need the data that are input in original software section FBP algorithm for reconstructing are proofreaied and correct processing, therefore the present invention needn't make amendment to the software section of original PET, directly adds one and proofreaies and correct the correction processing that the software performance module of handling just can realize effect of depth.Here need to prove that the parameter of the filtered back-projection method of image rectification front and back use needs consistent in the present invention.

In the present invention, because the distribution of imaging is relevant with crystal type and the detector configuration of the position of radioactive source, PET, for given crystal type and detector configuration, such as polygon or annular detector configuration, each source positions all has the image distribution of unique correspondence in the image, for technical scheme of the present invention is described better, use annular detector configuration to describe the present invention in detail here.

Figure 4 shows that the model of ring-type detector, comprise inside and outside two-layer crystal.Suppose that a point source S is positioned at the periphery in the detector image-forming visual field, the example that the internal layer crystal detects, image in the a-quadrant after the reconstruction, its image distribution is image_0, the imaging of a-quadrant is accurate imaging, approximate think do not have effect of depth (the crystalline effect of depth of internal layer is unconspicuous, and is negligible, so the approximate imaging of thinking accurate no effect of depth of the imaging of a-quadrant); And the example that crystal participation in the outside detects is imaged on the B zone after undergoing reconstruction, and its image distribution is image_doi, and the image in B zone is coarse image, mainly is subjected to the influence of effect of depth, and imaging is called the pseudo-shadow of DOI to visual field off-centring.

At the detector of above-mentioned configuration, the present invention obtains the image distribution probability data at described detector diverse location place in advance, utilizes specific software that the detector of concrete given configuration is simulated and obtains image distribution probability data.Wherein specific software can be GATE software, described image distribution probability data can adopt the sinogram form, be called for short the sino form, described sino data are forms of a kind of store data, it meets example with each encodes by the angle that meets line with to the radial distance at center, the visual field, is stored in the middle of the matrix.For a specific PET instrument, use GATE software and set up corresponding instrument models, simulate a certain position that a radioactive source is put in the visual field of detector, image data obtains this position sino_0 and sino_doi.Particularly, when described GATE software adopts hardware corrected DOI to the stratified method of crystal, be divided into two-layer uniformly to existing not stratified detector simulation, simulate the distributed data of each radioactive source imaging in the detector configuration, obtain accurate respectively and have the sino data of DOI effect, also promptly obtain the sino_0 data and the sino_doi data that distribute corresponding to the unique image_0 in each source positions place in the visual field and image_doi.After type, size and the arranged distribution of PET instrument detector crystal were determined, there just have unique sino_0 and sino_doi to distribute for the radioactive source of the unit strength of diverse location in the visual field to be corresponding with it.Sino_0 that obtains described here and sino_doi data have definite proportionate relationship, only depend on the position of radioactive source and the type and the structure of detector crystal.

Utilize the above-mentioned sino data that obtain to be applied in the original image that PET obtains then, and on basis, directly in the process of reconstructed image, carry out the correction of DOI effect based on employing FBP algorithm for reconstructing.The original image that wherein said PET obtains is to handle by the software section of the inner original FBP of the including algorithm for reconstructing of PET to obtain.Because the DOI effect is that hoop image internal ring also is the picture centre expansion outside image, therefore described correction is to cut the influence of its outer ring to DOI (B zone) effect of adjacent internal ring on polycyclic each ring image being divided into, and a DOI effect that cuts is converted on exact position that corresponding exact image is superimposed upon its outer ring (a-quadrant).Simultaneously since the DOI effect to internal extended, and the image of outer shroud can be similar to and thinks accurately not have the DOI effect, so, can know its influence by inference by outer shroud exact image, after internal ring cuts the image of outer shroud DOI effect, obtain the exact image of internal ring internal ring, know DOI influence again by inference to inferior internal ring image, obtain the exact image of time internal ring, inside successively, respectively encircled corresponding exact image and DOI effect image.For the image of each ring, deduct the DOI effect image of an outer shroud, and be that exact image compensates, thereby finish correction removing a fractional conversion.From the outer shroud to the internal ring, proofread and correct like this, obtain the image behind the correction of a final proof by ring.

Specifically describe the principle that the described FBP of utilization algorithm for reconstructing carries out above-mentioned DOI effect calibration below.

As seen from the above description, when adopting the FBP algorithm for reconstructing to carry out image rectification, the image value of internal ring will deduct the DOI effect image of outer shroud to internal ring, and this part a DOI effect that deducts is converted to the exact image post-compensation in the image of outer shroud, go down so successively, until the image of having proofreaied and correct all rings, therefore concerning the image of any ring, it rebuilds the process of proofreading and correct the DOI effect is exactly two parts: one is to remove part, to this ring image, cut the image of its outer shroud to its DOI influence; Another is a compensated part, is the image transitions that its internal ring is produced the DOI influence accurate image, adds on this ring image.

Suppose to be defined as image with not passed through the gauged original image of DOI after the reconstruction of FBP algorithm.This original image can be regarded as the pixel stack of N*N, and each image pixel size is x, is the center with the center of detector, sets up polar coordinate system, and image radius r ∈ (0, r _N], θ ∈ (0,2 π).This original image can be regarded as many varying strength A that are positioned at diverse location, and (intensity of radioactive source is equal to after the radioactive source imaging gray value of pixel in the image here for r, the stack of radioactive source imaging θ).Image can expand into so

$\mathrm{image}=\underset{r,\mathrm{\θ}}{\mathrm{\Σ}}(A(r,\mathrm{\θ})\×\mathrm{image}(r,\mathrm{\θ}))---\left(1\right)$

Here (r is to be positioned at (r, the imaging distribution of unit strength radioactive source θ) θ) to image.

And the imaging distribution image of each position (r, θ) can regard the stack of image_0 (not having the DOI effect) and image_doi (the DOI effect is arranged) as, wherein image_0 and image_doi can obtain by obtained sino Probability Distribution date processing by simulation, therefore, concerning each position, its image distribution is:

image(r _j，θ)＝image_0(r _j，θ)+image_doi(r _j-1，θ)(j＝0，1...N)(2)

Here r ₀=0, r _NMaximum radius value for image in the vision detector.

Formula (1) and formula (2) combine, and the like, obtain:

$\mathrm{image}=\underset{j,\mathrm{\θ}}{\mathrm{\Σ}}(A(r_j,\mathrm{\θ})\×(\mathrm{image}\_0(r_j,\mathrm{\θ})+\mathrm{image}\_\mathrm{doi}(r_{j-1},\mathrm{\θ})))---\left(3\right)$

$=\underset{\mathrm{\θ}}{\mathrm{\Σ}}\underset{\‾}{(...+A(r_j,\mathrm{\θ})\×\mathrm{image}\_0(r_j,\mathrm{\θ})}+(\underset{\‾}{A(r_j,\mathrm{\θ})\×\mathrm{image}\_\mathrm{doi}(r_{j-1},\mathrm{\θ})+A(r_{j-1},\mathrm{\θ})\×\mathrm{image}\_0(r_{j-1},\mathrm{\θ}))}+...)$

And in fact the original image that observes of PET is:

$\mathrm{image}=\underset{\mathrm{\θ}}{\mathrm{\Σ}}\underset{\‾}{(...+(A(r_j,\mathrm{\θ})\×\mathrm{image}\_0(r_j,\mathrm{\θ})}+\underset{\‾}{B(r_{j-1},\mathrm{\θ})\×\mathrm{image}\_0(r_{j-1},\mathrm{\θ})}+...)---\left(4\right)$

Because formula 3 is analyses that each position image value is carried out, formula 4 is images that we directly observe, and in theory, directly observes image and should be equal to the image that our analyze, and therefore can obtain:

$\underset{\mathrm{\θ}}{\mathrm{\Σ}}A(r_{j-1},\mathrm{\θ})\×\mathrm{image}\_0(r_{j-1},\mathrm{\θ})=\underset{\mathrm{\θ}}{\mathrm{\Σ}}(B(r_{j-1},\mathrm{\θ})\×\mathrm{image}\_0(r_{j-1},\mathrm{\θ})-A(r_j,\mathrm{\θ})\×\mathrm{image}\_\mathrm{doi}(r_{j-1},\mathrm{\θ}))---\left(5\right)$

According to formula (5), can know for radius r _J-1Image need remove one the ring radius be r _jThe DOI image, the principle of the removal of DOI effect calibration process that Here it is part.

Be r for radius equally _J-1The radioactive source at ring place is also to next ring r _J-2The image at place produces the DOI effect, and need compensate to radius to the reconstruction of the DOI effect of this part is r _J-1The ring place, that is to say for each ring, should comprise the removal of a ring DOI effect, also to comprise the compensation that next ring is produced the DOI effect, different with the removal part is, compensation described here be not simply DOI effect image overlay that next ring is produced to the exact image of this ring, but the image approximate of DOI effects is converted to exact image under the same quantity of data, compensate on this ring image.

For the sino data of each position, its sino_0 and sino_doi have certain proportionate relationship, also promptly as shown in the formula: $C(r_j,\mathrm{\θ})=\frac{\mathrm{\Σ}(\sin o\_\mathrm{doi}(r_j,\mathrm{\θ}))}{\mathrm{\Σ}(\sin o\_0(r_j,\mathrm{\θ}))},$ Therefore, according to C (r _j, proportionate relationship θ), the sino_doi of each position can convert into corresponding precise information C (r _j, θ) * sino_0 (r _j, θ); For (r _j, θ) position produces the image A (r of DOI effect _j, θ) * image_doi (r _j, θ), the sino data that obtain by simulation and the corresponding relation of image image, corresponding sino data are A (r _j, θ) * sino_doi (r _j, θ), so this part data can be regarded the precise information A (r as same quantity of data as _j, θ) * C (r _j, θ) * sino_0 (r _j, θ), correspondingly, this part A (r _j, θ) * image_doi (r _j, image θ) can be converted to the A (r under the same quantity of data _j, θ) * C (r _j, θ) * image_0 (r _j, θ), then the A (r after the conversion _j, θ) * C (r _j, θ) * image_0 (r _j, image value θ) compensates on this ring image.

Proofread and correct to the inner ring that pursues of image by the image periphery like this, can obtain

$\mathrm{image}=\underset{j,\mathrm{\θ}}{\mathrm{\Σ}}((A(r_j,\mathrm{\θ})+C(r_j,\mathrm{\θ}))\×\mathrm{image}\_0(r_j,\mathrm{\θ})))(j=\mathrm{0,1}...N)---\left(6\right)$

According to above-mentioned principle, technical scheme of the present invention is described in detail below in conjunction with an embodiment.

At first adopt the GATE software simulation to set up the detector model, the crystal bar length of detector is constant, the principle of stratification of copying hardware, crystal is divided into two-layer, the gamma photon that it is 511keV that radioactive source discharges two energy, two photon exit directions are opposite, are approximately 180 degree.A pair of gamma photon like this is called one and meets example, and the example that two photons are all detected by the internal layer crystal is called as the internal layer example, layer data sino_0 in being designated as with the form of sino, and this situation DOI effect is not obvious; The example of other situations, detected by internal layer as two one of photons possibilities, the outer crystal of quilt detects, and also might two photons be all detected by outer crystal and all thinks the tangible example of DOI effect, and the example of both of these case is designated as outer layer data sino_doi.To each locations of pixels among the figure, be the radioactive source of x with GATE software simulation size, obtain internal layer example sino_0, sino_doi data.

Handle sino_0 respectively with the FBP algorithm for reconstructing of PET software section then and the sino_doi data obtain image image_0 and image_doi, and make sino_0+sino_doi=sino_all, utilize the FBP algorithm for reconstructing to handle sino_all and obtain image image (r, θ), and image image (r, θ) just radioactive source is located at position (r, θ) locating the entire image that obtains distributes, wherein the position (r, the grey scale pixel value of θ) locating be A (r, θ).Sino_0 wherein as herein described, sino_doi, sino_all, image_0, image_doi, image (r, θ) all be sino and view data after the normalized processing, so-called normalized, be to primary sino and view data divided by A (r, θ).

Because annular detector can be similar to and be seen as is that circle is symmetric, does not need so each pixel coordinate is made analogue measurement.Only need to measure one group of different r back-to-back radioactive source value, that be of a size of x of θ=0, to its processing and storage sino and view data, and the data of other different θ according to the symmetry of circle, can and be inverted the sino data by translation and obtain.Such as obtaining radioactive source, simulation is positioned at (r, θ=0) sino data, obtain radioactive source and be positioned at (r, θ ≠ 0) sino data, if θ＜180 degree, with the separation delta θ of θ divided by θ in the sino data, its merchant rounds and is n, the circulation n row sino data that move right, and the row that first in the sino matrix is listed as the n row is inverted.If θ＞180 degree, divided by separation delta θ, its merchant rounds and is n with θ-180, and circulation is moved to the left n row sino data, and the row that is listed as the n row last in the sino matrix is inverted.Circulation is moved and to be obtained the sino data with inversion and be radioactive source and be positioned at identical r, the sino data of different θ like this.

The sino data at the diverse location place that above-mentioned simulation is obtained and image data storage are in the PET instrument, in order to using in carrying out following gauged process.For specific detector model, need simulate above-mentioned sino data and image distribution data again, and be stored in the PET instrument.

Next, as shown in Figure 5, to original image according to carrying out the correction of following steps to the order of innermost ring from outer shroud:

Detect step, for selected ring, whether the image that detects this arbitrary position of selected ring is that the radioactive source imaging of position thus forms, and is then, changes the correction data obtaining step over to, otherwise continues to detect the next position image of this selected ring;

The correction data obtaining step, extract the sino_doi and the sino_0 of this position, according to the proportionate relationship of this position C=(∑ sino_doi)/(∑ sino_0), obtain to be used to compensate this selected ring produces effect of depth to next ring compensation distributed data C*sino_0 simultaneously.Change the S21 step then over to, up to the compensation distributed data that has obtained all positions of described selected ring;

The cumulative data step, sino_doi of this each position of selected ring added up obtains sino_d, and the compensation distributed data of described each position added up obtains sino_a;

The correction image step, the sino_d data of handling this selected ring by filtered back projection's FBP algorithm for reconstructing obtain image value image_d, and handle described sino_a data and obtain image value image_a, proofreading and correct image value of handling behind this selected ring is the image value subtracted image value image_d after proofreading and correct last time and adds image value image_a, change over to then and detect the circulation correction processing that step is carried out next ring image, up to handling all images.

Be described in detail with regard to each step below.

(1) detects step

Because one at (r _j, θ) the image_0 imaging of radioactive source distribution is with (r in the image _j, θ) pixel compartments of this coordinate position is that the round speckle of of center distributes, and that is to say the image after each radioactive source imaging, does not just take a pixel compartments, its imaging meeting influence in the image grey scale pixel value on every side.So the value of any one pixel compartments in the image is just needed to judge that it has the radioactive source imaging to form in this position, still form by the radioactive source imaging expansion of other positions or noise.

Detection method is: suppose in the image at (r _j, θ) pixel value of this position is A _jIf there is radioactive source in this position and imaging just should comprise A so in image so _j* image_0 (r _j, distribution θ), because influence of noise, as long as require image to comprise A here _j* image_0 (r _j, θ) major part of Fen Buing just thinks that there is radioactive source in this position.

The concrete operations step is as follows:

1) in wanting the image image of analyzing and processing, finds out gray value greater than A _j* the pixel of ε, its coordinate is index, extracts image image and (r _j, θ) the index coordinate corresponding pixel value among the image image_0 that obtains is simulated in the position, is labeled as image (index) and A respectively _j* image_0 (index), both make difference and obtain I, I (index)=image (index)-A _j* image_0 (index).

2) have only on the principle coordinate be all I values of index greater than 0, illustrate that then image comprises A _j* image_0 (r _j, θ) distributing, but because effect of noise thinks during processing that image comprises major part just, the number proportion that just departs from 0 very big element among the I is very little.This can be weighed with empirical parameter δ, ζ, asks I less than-δ * A _jNumber be labeled as N; If less than the ζ of index number, then being similar to, N thinks that the image of this position is real radioactive source image, otherwise, think that then original image does not have radioactive source in this position.

Wherein be used for checking empirical parameter ε, δ, the ζ of radioactive source existence here, be the empirical parameter that obtains at different detector configurations and concrete experiment condition, the value of this empirical parameter satisfies condition to the accuracy that makes detection approaches practical situation, and best here value all is 0.1.

The process of described detection must be done before proofread and correct each position, just, will detect earlier in each position, proofreaied and correct again, carried out the processing of next position afterwards.

(2) correction data obtaining step and cumulative data step

According to above-mentioned correction principle narration, in carrying out trimming process, comprise the removal and the compensation of image DOI effect, therefore, in realization, need know that current ring image to be processed produces the imaging distribution sino data of DOI effect to next ring image, and need know to convert to become to be used for and compensate the sino offset data of current ring DOI effect.

Because proofreading and correct is the process of a circulation recurrence, we earlier begin to proofread and correct from the image of any angle position of outer shroud and handle, proofreaied and correct after the image of all angular position of outer shroud, proofread and correct next the ring image handle outer shroud next-door neighbour again, and the like, up to having proofreaied and correct all images.

At first need to find out outer shroud image, just obtain finding out in the image image value greater than 0 and the outer shroud image of radius maximum, suppose that the radius of above-mentioned outer shroud image is r above-mentioned with image value and radius maximum _N, M some pixel arranged on this ring radius, the angle value is θ=θ ₁. θ ₂..., θ _M, at (r=r _N, θ ₁) the gray value A that locates _N(r=r _N, θ ₁), obtain in the above-mentioned steps that radioactive source is positioned at (r=r under the unit strength _N, θ ₁) sino_0 (r=r that locates _N, θ ₁) and sino_doi (r=r _N, θ ₁) data, and according to sino_0 (r=r _N, θ ₁) obtain accurate image_0 (r=r after the reconstruction _N, θ ₁).

The parameter of initialization intermediate variable, sino_d=0, sino_a=0, image_t=image.Adopt sino_d to represent the sino distributed data of image generation DOI effect in the single treatment, sino_a represents will compensate in the single treatment sino distributed data of processing, and image_t is expressed as in original image remainingly after the single treatment does not have a gauged image section.The single treatment here, the only single treatment that the image of different angles under the radius is finished.

Image for outer shroud radioactive source, we at first will obtain the imaging distributed data of all position radioactive sources on the outer shroud, then this institute is useful on gauged imaging distributed data and adds up, and then carry out FBP and rebuild and to obtain being used to the image distribution that compensates and remove.Proofreading and correct process points (r=r _N, θ ₁) before the image located, carry out the above-mentioned detection of mentioning earlier and handle, if detect the imaging that this position is a radioactive source, just the sino data accumulation of the generation DOI effect of this position to initializaing variable, also be variable sino_d=sino_d+A _N* sino_doi (r=r _N, θ ₁), and get C=(∑ sino_doi (r=r _N, θ ₁))/(∑ sino_0 (r=r _N, θ ₁)) (the sino data are matrix distribution data, the summation here is all numerical value summations in the matrix), then according to above-mentioned compensation principle, the sino data that produce the DOI effect are converted into accurate sino data under the same quantity of data, obtain sino_a=sino_a+A _N* C*sino_0 (r=r _N, θ ₁), the residual image after then this point source is handled is image_t=image_t-image_0 (r=r _N, θ ₁) * A _N(r=r _N, θ ₁).

Detect and handle the next angle θ on this ring then ₂The imaging of radioactive source is all handled up to the radioactive source of this ring being gone up all positions, adds up successively at last to obtain being used for rebuilding and carries out gauged imaging distributed data sino_d and sino_a.

(3) correction image step

Sino_d data according to adding up and obtain on the above-mentioned outer shroud obtain image_doi by the FBP algorithm for reconstructing, and image_doi represents r=r here _NThe DOI effect image that the image of position produces next ring; Obtain image_add with obtaining the sino_a reconstruction, image_add indicates to compensate to r=r here _NThe compensating images of position distributes.

The image of having proofreaied and correct behind the outer shroud is image_c=image-image_doi+image_add, just, deducts DOI effect image_doi part in the original image, and that DOI fractional conversion is partly compensated on the image for approximate accurate image_add.The residual image of handling after the outer shroud image is image_t=image_t-image_doi.

Proofreading and correct next radius then is r _N-1The internal ring image, and the sino_d=0 that resets, sino_a=0 handles residual image image_t, changes over to and detects in the step, carries out aforesaid operations successively, upgrades image_c and image_t, up to image_t=0.The image image_c that obtains at last is the image after proofreading and correct.

Above step is that to carry out DOI at entire image gauged, if only the C in the image is partly proofreaied and correct, that needs to delimit certain zone, as shown in Figure 6, grid is an original image, only need revise at interior selection area D inclusion region C and get final product, the remainder assignment is 0, obtains region of interest area image image_ROI.The selection of image_ROI comprises the zone of correction portion, and extends to the outer shroud of image in radial direction.

Adopt experiment to prove the present invention's effect of depth in the removal of images at an easy rate below.

In experiment, adopt the experimental apparatus of e-plus PET, this detector is axially having 32 rings, and each encircles 16 modules, and 16 LYSO crystal bars are arranged in each module.This crystal bar is of a size of 2*2*10mm ³Adopt the Phantom model to be used for the model of Measurement Resolution, the poly cylindrical radius that is configured to of this model is 5cm, thickness is 2.1cm, and six groups cylindric aperture is arranged, and the diameter of six groups of apertures is respectively 1.4,1.6,1.9,2.2,2.5 and 3.0mm, every group of all hexagonal arrangement of aperture, and the center distance in adjacent two holes is the twice of diameter.For on the same group aperture not, the minimum distance at aperture center position Phantom center all is 8mm.Injection of radioactive substances in aperture is put into the imaging of PET instrument.Whether can identification in experiment with adjacent two small bore columns sources in the image, be used for estimating image resolution ratio.

Experimental implementation condition: phantom is put in departs from 1.6cm place, center, the visual field, and diameter is to fill identical activity in 4 apertures of a radially row of 1.9mm in phantom ¹⁸The FDG medicine.Actual measurement obtains 4 source positions, and (X, Y) coordinate is as follows: (33.15,2.5) mm, (36.95,2.5) mm, (40.75,2.5) mm, (44.55,2.5) mm.

Experimental apparatus parameter: adopt the energy window of 350-650keV, adopt the time window of 12.5ns.

The image reconstruction parameter: the parameter of FBP algorithm for reconstructing is: filter function is the Hamming window; Interpolation method is arest neighbors interpolation (nearest_neighbor); Cut-off frequency is d=0.95.

Experimental result shows: before the correction, isolating radioactive source has been linked to be wire.After above-mentioned DOI bearing calibration correction, different radioactive sources are separated, and image has radially been differentiated significant improvement.And need not in the place that does not have image to handle, convenient, flexible, save time.

Though the above-mentioned example of executing has been described the present invention, should be appreciated that used term is explanation and exemplary and nonrestrictive term.The spirit or the essence that do not break away from invention because the present invention can specifically implement in a variety of forms, so be to be understood that, the foregoing description is not limited to any aforesaid details, and should be in the spirit and scope that claim limited of enclosing explain widely, therefore fall into whole variations in claim or its equivalent scope and remodeling and all should be the claim of enclosing and contain.

Claims (8)

1. a method for correcting image depth effect is used to proofread and correct the position emissron tomography instrument by the original image that filtered back projection's algorithm for reconstructing obtains, and it is characterized in that, said method comprising the steps of:

S1: obtain the interior every bit position of described position emissron tomography instrument vision detector scope and be used for the Probability Distribution data that imaging distributes and rebuilds;

S2: utilize described filtered back projection algorithm for reconstructing to handle described Probability Distribution data, according to described original image being carried out the removal of effect of depth image to the order of innermost ring from outer shroud, and be that corresponding exact image compensates described original image, thereby finish the correction of effect of depth to a fractional conversion of removing.

2. bearing calibration according to claim 1 is characterized in that, described step S2 may further comprise the steps further:

S21: for the selected ring of present image, whether the image that detects the arbitrary position of described selected ring is that the radioactive source imaging by described position forms, and is then, changes the S22 step over to, otherwise continues to detect the next position image of described selected ring;

S22: the Probability Distribution data of extracting described position, described Probability Distribution data comprise internal layer data sino_0 and outer layer data sino_doi, while is according to the proportionate relationship of described position C=(∑ sino_doi)/(∑ sino_0), acquisition is used to compensate this selected ring produces effect of depth to next ring compensation distributed data C*A*sino_0, change the S21 step then over to, up to the compensation distributed data that has obtained all positions of described selected ring, wherein said A is the pixel value of described position;

S23: the sino_doi of each position of described selected ring added up obtains sino_d, and the compensation distributed data of described each position added up obtains sino_a;

S24: the sino_d data of handling described selected ring by filtered back projection's algorithm for reconstructing obtain image image_d, and handle described sino_a data and obtain image image_a, having proofreaied and correct this image of selecting after encircling is described present image value subtracted image image_d and adds epigraph image_a, change the S21 step then over to and proofread and correct next ring image of processing, described present image is the image behind the intact last ring of described correction, up to handling all original images.

3. bearing calibration according to claim 2 is characterized in that, described original image is the part in the entire image that obtains of described position emissron tomography instrument.

4. bearing calibration according to claim 3 is characterized in that, described original image is an image value greater than zero image, is zero to other image section assignment of not proofreading and correct processing in the described entire image.

5. bearing calibration according to claim 2, it is characterized in that, detection method in described step S21 is: the data sino_0 that adopts filtered back projection's algorithm for reconstructing to handle described position obtains image distribution image_0, if the image distribution of original image in described position comprises the image distribution of image_0, think that then the image of position described in the described original image is that radioactive source by described position forms.

6. according to any described bearing calibration of claim 1-5, it is characterized in that the described sinusoidal histogrammic data format of Probability Distribution The data that is used for imaging distribution reconstruction.

7. according to any described bearing calibration of claim 1-5, it is characterized in that, the every bit position is used for the Probability Distribution data that imaging distributes and rebuilds in the described described vision detector scope of obtaining in advance, comes analog prober to obtain by adopting GATE software.

8. bearing calibration according to claim 6 is characterized in that, described GATE software simulation crystal divides inside and outside two-layer method to obtain described Probability Distribution data.

Images