JP2016187462A - Medical image processor and medical image processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a medical image processor capable of correcting a medical image to accurately track an anatomical change of an object.SOLUTION: A medical image processor 1 includes: an area setting part 13 for setting a first area corresponding to a part of a first image, and a second area corresponding to a part of a second image collected earlier than the first image; a frequency distribution generation part 15 for generating a first area frequency distribution on the basis of a plurality of pixel values in the first area, and a second area frequency distribution on the basis of a plurality of pixel values in the second area; and a correction part 17 for correcting at least one of the first frequency distribution corresponding to the first image and the second frequency distribution corresponding to the second image with the first area frequency distribution, the second area frequency distribution or a predetermined frequency distribution as a reference.SELECTED DRAWING: Figure 1

Description

  Embodiments described herein relate generally to a medical image processing apparatus and a medical image processing method for correcting a medical image.

  In Magnetic Resonance Imaging (MRI), three-dimensional volume data of the brain such as T1W, T2W, etc., using a statistical analysis tool such as SPM (Statistical Parametric Mapping) (Voxel-BasedMedVorbMemVorbMorV) There is a field called. In VBM, anatomical standardization of the individual brain into the standard brain is performed after alignment, but generally, strong smoothing is applied to eliminate minute changes. That is, in VBM, anatomical standardization (registration) to the standard brain is executed by intentionally blurring the target image.

  In the field of VBM, it has been applied to dementia and neuropsychiatric diseases that require tracking of minute changes in form with time. For example, in the diagnosis of dementia such as Alzheimer (AD), the thickness of the cortex is in the order of sub mm, the brain volume of the hippocampus is in the order of several percent, and MRI is used to measure the minute amount over time. Such changes need to be tracked from time to time in units of one to several months, from several years to life.

  An analysis method similar to VBM has been widely applied not only to structural data, but also to functional data such as diffusion images, functional MRI (fMRI), and cerebral blood flow (Arter Spin Labeling: ASL). . The modalities are not limited to MRI apparatuses, but are not limited to single photon emission computed tomography (SPECT) apparatus, positron emission computed tomography (PET) apparatus, and X-ray CT (X-ray CT). ) There are many applications in equipment.

  For chest X-ray images, a method has been put to practical use in which a time difference is performed using an image captured every certain period to make it easier to grasp the change from the previous time. In this case, using information such as bones, correction of the patient imaging position in the depth direction peculiar to the transmission image is performed in addition to the two-dimensional imaging position in the plane. In addition, even in an imaging apparatus capable of acquiring cross-sectional images such as X-ray CT, SPECT, PET, and MRI, re-slicing is performed by recutting the previous cross section.

  However, in a plurality of medical images having different imaging times, if there is a difference between apparatuses or a filter or a matrix size related to imaging and image processing, it is difficult to track a correct temporal change due to a temporal change factor derived from a patient. Further, even when the same apparatus conditions (imaging conditions) are used, it is necessary to perform image interpolation when correcting a change in imaging position. Even if the filters are equivalent, for example, an effect such as blur may appear due to a difference in the number of interpolations, and there is a problem that these influences are not currently considered.

Japanese Patent No. 4022587 Japanese Patent No. 5532730 JP 2014-042684 A JP 2010-046103 A

  An object of the present invention is to provide a medical image processing apparatus and a medical image processing method capable of correcting a medical image so that an anatomical change of an object can be accurately tracked.

  The medical image processing apparatus according to this embodiment sets a first area corresponding to a part of the first image and a second area corresponding to a part of the second image collected in the past from the first image. A setting unit; and a frequency distribution generation unit that generates a first region frequency distribution based on a plurality of pixel values in the first region and generates a second region frequency distribution based on the plurality of pixel values in the second region; A first frequency distribution corresponding to the first image and a second frequency distribution corresponding to the second image on the basis of the first region frequency distribution, the second region frequency distribution, or a predetermined frequency distribution. And a correction unit that corrects at least one of them.

FIG. 1 is a configuration diagram showing the configuration of the medical image processing apparatus according to the first embodiment. FIG. 2 is a diagram illustrating an example of setting the first region or the second region in the first interpolation image and the second interpolation image according to the first embodiment. FIG. 3 is a diagram illustrating an example of a change (noise profile) of a pixel value with respect to a position (for example, an air portion) in the first region or the second region according to the first embodiment. FIG. 4 relates to the first embodiment, the power spectrum corresponding to the first domain frequency distribution generated by using the inverse Fourier transform of the noise profile, the power spectrum corresponding to the second domain frequency distribution, and their powers. It is a figure which shows an example of the 1st smooth spectrum and the 2nd smooth spectrum which each smoothed the spectrum. FIG. 5 is a diagram illustrating an example of a conversion function for converting the first smoothed spectrum into the second smoothed spectrum according to the first embodiment. FIG. 6 is a diagram illustrating an example of a power spectrum corresponding to the first frequency distribution corrected using the conversion function according to the first embodiment. FIG. 7 is a flowchart illustrating an example of a procedure of frequency distribution correction processing according to the first embodiment. FIG. 8 is a configuration diagram showing the configuration of the medical image processing apparatus according to the second embodiment. FIG. 9 is a diagram illustrating an example of setting the first region or the second region in the first interpolation image and the second interpolation image according to the second embodiment. FIG. 10 is a diagram illustrating an example of a change in pixel value (actual signal profile) on the line segment ee ′ in FIG. 9 according to the second embodiment. FIG. 11 is a diagram illustrating an example of a change (time-invariant profile) of a pixel value on the line segment dd ′ in FIG. 9 or the extraction portion (ff ′) in FIG. 10 according to the second embodiment. FIG. 12 is a diagram illustrating an example of the first normalized frequency distribution and the second normalized frequency distribution generated by normalizing the absolute value of the inverse Fourier transform of the time-invariant profile according to the second embodiment. . FIG. 13 is a diagram illustrating an example of a conversion function for converting the first normalized frequency distribution into the second normalized frequency distribution according to the second embodiment. FIG. 14 is a diagram illustrating an example of the absolute value of the first frequency distribution corrected by using the conversion function according to the second embodiment. FIG. 15 is a flowchart illustrating an example of a procedure of frequency distribution correction processing according to the second embodiment.

  Hereinafter, a medical image processing apparatus according to the present embodiment will be described with reference to the drawings. In the following description, components having substantially the same function and configuration are denoted by the same reference numerals, and redundant description will be given when necessary.

(First embodiment)
FIG. 1 is a block diagram illustrating a configuration of a medical image processing apparatus 1 according to the first embodiment. The medical image processing apparatus 1 includes an interpolation processing unit 11, a region setting unit 13, a frequency distribution generation unit 15, a correction unit 17, an image generation unit 19, a display unit 21, an interface unit 23, and an input unit 25. And a storage unit 27 and a control unit 29. A in FIG. 1 represents a network connected to various external modalities, medical image storage devices, and the like via the interface unit 23 in the medical image processing apparatus 1.

  When the first image and the second image are selected according to an instruction from the operator via the input unit 25, the interpolation processing unit 11 performs the first image and the second image (using an interpolation function). Perform interpolation processing. The interpolation processing is, for example, unification of voxel (or pixel) size, spatial position correction, alignment, and the like in the first image and the second image. The first image is a medical image (latest or current medical image) to be compared with the second image. The second image is a medical image that serves as a reference for the first image. Specifically, the second image is usually compared with the current (first image), and the medical image acquired in the past than the first image or the volume data collected in the past is reproduced with respect to the reference cross section. It is a sliced (re-sliced) medical image (for example, an MPR (Multi-Planar Reconstruction) image).

  The first image and the second image are medical images having the same anatomical region related to the subject. Hereinafter, the first image and the second image will be described as brain images. The first image and the second image are not limited to brain images, and may be images having arbitrary anatomical regions. The first image and the second image are, for example, a medical image having a region corresponding to cancer, a medical image related to osteoporosis, a medical image used as an evaluation (infiltration degree) of cancer progression, a medical image related to a follow-up examination, and the like. is there. The first image and the second image are medical images generated with the same modality.

  Specifically, after the selection of the second image, the interpolation processing unit 11 determines an image condition serving as a reference (Base) based on the second image, for example. The image conditions include, for example, a matrix size, a voxel size based on an imaging field of view (FOV), an imaging position of a patient, and the like. In addition, the image condition used as a reference | standard may be determined based on the incidental information of a 1st image. Further, the reference image condition may be determined based on the incidental information of the second image and the incidental information of the first image.

  For example, the interpolation processing for the spatial position difference between the first image and the second image is correction of the position in the plane if the first image and the second image are two-dimensional images, and the first image and the second image are three-dimensional images. This corresponds to correction of the slice cross section. In addition, since the processing for the difference in voxel size and spatial position is “interpolation”, it is desirable to perform similar processing (interpolation processing using a similar interpolation function) on the first image and the second image.

  In addition, it is important to perform interpolation processing when aligning the target image (first image) with respect to the reference image (second image) using a technique with as little frequency degradation as possible with high accuracy. It is necessary to determine the conversion filter characteristics in consideration of the degradation of the power spectrum at the time.

  The region setting unit 13 sets a first region corresponding to a part of the first image after interpolation processing (hereinafter referred to as a first interpolation image). The region setting unit 13 sets a second region corresponding to the second image after interpolation processing (hereinafter referred to as a second interpolation image). The first region is a spatially uniform region in the first interpolation image. The second region is spatially uniform in the second interpolation image.

  That is, the first region and the second region are, for example, regions related to a noise portion, and are regions related to a non-signal portion (air region) and a structure having no structure (non-structure) and spatially uniform. . The first region and the second region are, for example, a region corresponding to air and a region corresponding to the ventricle in a brain image. Hereinafter, the first region and the second region are regions that are non-structured and spatially uniform (no signal portion, noise portion), that is, an air region. The case where the first region and the second region are regions that change little over time (regions that do not change over time) will be described in detail in the second embodiment.

  FIG. 2 is a diagram illustrating an example of setting the first region or the second region in the first interpolation image and the second interpolation image. In FIG. 2, a indicates an air region in the interpolation image. In FIG. 2, b indicates a bone region in the interpolation image. In FIG. 2, c indicates a scalp region in the interpolated image. GM in FIG. 2 indicates an area of gray matter in the interpolated image. WM in FIG. 2 indicates a white matter area in the interpolated image. CSF in FIG. 2 indicates a region having cerebrospinal fluid (CSF) in the interpolation image. In FIG. 2, the CSF at the center of the image that is bilaterally symmetric corresponds to the ventricle. N in FIG. 2 corresponds to an area where the first area and the second area are set, for example.

  The region setting unit 13 sets the first region based on the pixel value of the first interpolation image, and sets the second region based on the pixel value of the second interpolation image. For example, the region setting unit 13 sets the first region to the first interpolation image based on threshold values for a plurality of pixel values in the first interpolation image. The region setting unit 13 sets the second region in the second interpolation image based on threshold values for a plurality of pixel values in the second interpolation image. The threshold value is, for example, a pixel value from which air can be extracted. This \ threshold value is stored in the storage unit 27.

  Hereinafter, in order to simplify the description, it is assumed that the shapes of the first region and the second region are rectangular. Note that the shapes of the first region and the second region are not limited to a rectangle, and may be any shape such as a circle or an ellipse. The first region and the second region may be an air region or a CSF region. When the first region or the second region cannot be set as the air region, the region setting unit 13 may set the first region and the second region as the CSF region, for example.

  The size of the first region and the size of the second region are preferably the same size. Note that the size of the first region may be different from the size of the second region. The shapes and sizes of the first area and the second area are set in advance. Note that the shape and size of the first region and the second region can be appropriately changed according to an instruction from the operator via the input unit 25. The first area and the second area may be set according to an instruction from the operator via the input unit 25. The size of the first region and the size of the second region have, for example, several hundred pixels.

  When the first interpolation image and the second interpolation image are amplitude images, the signal-to-noise ratio (Signal-Noise Ratio: SNR) is sufficiently large (SNR> 5). Therefore, a plurality of non-air regions (signal portions) The signal value has a Gaussian distribution. On the other hand, the plurality of signal values (noise) in the air region (no-signal portion) have, for example, a non-negative (non-negative) Rician distribution. For this reason, in the processing after setting the first region (and the second region), the signal part and the non-signal part are handled separately, or the first interpolation image (second interpolation image) to the first interpolation image (second interpolation). In the difference image obtained by subtracting the DC (Direct Current) component (average pixel value) of the interpolated image (after removing the DC component), the first region (and the second region) is converted into an absolute value (magnitude). It is necessary to convert the no-signal part in to Lycian noise.

Hereinafter, a plurality of pixel values respectively corresponding to a plurality of pixels included in the first region set in the first interpolation image are defined as I _orig (x). In addition, a plurality of pixel values respectively corresponding to a plurality of pixels included in the second region set in the second interpolation image are set to I _base (x). In order to simplify the description, it is assumed that the positions of the plurality of pixels included in the first area and the positions of the plurality of pixels included in the second area are the same. At this time, the first interpolation image and the second interpolation image are registered in advance by the interpolation processing unit 11. Further, when one of the first area and the second area is set, the other may be automatically set. An argument x in I _orig (x) and I _base (x) indicates the position of a pixel included in the first area and the second area. Further, the range including the argument x is the same. In general, when the first area and the second area are set at different positions and different sizes, the arguments are defined by different characters (for example, I _orig (x) and I _base (y)). The range of arguments is also different.

  FIG. 3 is a diagram illustrating an example of a change (noise profile) of a pixel value with respect to a position (for example, an air portion) in the first region (or the second region). The vertical axis in FIG. 3 indicates the pixel value. The ave on the vertical axis in FIG. 3 corresponds to the average pixel value (DC component) of the first area (or the second area).

  When the first region and the second region are two-dimensional, the noise profile corresponding to FIG. 3 is a three-dimensional graph (the x-axis and y-axis are pixel positions, and the z-axis is noise magnitude). When the first image and the second image are volume data, the noise profile corresponding to FIG. 3 is a four-dimensional graph (the x-axis, y-axis, and z-axis are pixel positions, and the w-axis is the magnitude of noise. Is).

  The frequency distribution generation unit 15 generates a first power spectrum indicating the square of the magnitude of the frequency distribution in the first region based on the plurality of pixel values included in the first region, and the plurality of pixel values in the second region. To generate a second power spectrum indicating the square of the magnitude of the frequency distribution in the second region. Specifically, the frequency distribution generation unit 15 performs inverse Fourier transform on a plurality of pixel values included in each of the first region and the second region. Next, the frequency distribution generation unit 15 generates the first power spectrum and the second power spectrum by applying the absolute value and the square to the result of the inverse Fourier transform, respectively.

When the first power spectrum is expressed as P _orig (f) and the second power spectrum is expressed as P _base (f), the calculation executed in the frequency distribution generation unit 15 can be expressed as follows, for example. Here, f is a frequency and an argument in the first power spectrum and the second power spectrum.
P _orig (f) = | IFT [I _orig (x)] | ²
P _base (f) = | IFT [I _base (x)] | ² (1)
The IFT in the above equation indicates the inverse Fourier transform.

  That is, the frequency distribution generation unit 15 performs a two-dimensional or three-dimensional inverse Fourier transform on a plurality of pixel values (image data) in each of the first region and the second region according to the dimension of the first (interpolated) image. Execute. A first power spectrum and a second power spectrum respectively corresponding to the first region and the second region are generated by the absolute value calculation and the square calculation for the inverse Fourier transform.

Generally, the scaling of pixel values in the first image and the second image is different depending on the gain at the time of imaging. For this reason, the frequency distribution generation unit 15 normalizes the first power spectrum and the second power spectrum with the maximum value (usually a DC (f = 0) component) of the reference power spectrum. In order to simplify the description, it is assumed that the reference power spectrum is the maximum value P _base (0) of the second power spectrum. A case where the reference power spectrum is the first power spectrum and a case where the power spectrum is a power spectrum corresponding to a predetermined frequency distribution will be described later in a modification.

Specifically, the frequency distribution generation unit 15 normalizes the first power spectrum by dividing the first power spectrum P _orig (f) by the maximum value P _base (0) of the second power spectrum. Further, the frequency distribution generation unit 15 normalizes the second power spectrum by dividing the second power spectrum P _base (f) by the maximum value P _base (0) of the second power spectrum.

That is, the normalized first power spectrum Pn _orig (f) is calculated as P _orig (f) / P _base (0). The normalized second power spectrum Pn _base (f) is calculated as P _base (f) / P _base (0). The frequency distribution generator 15 calculates a normalization coefficient α. For example, the normalization coefficient α is calculated as in the following equation.
α = P _base (0) / P _orig (0) (2)
When the first region and the second region have signal components other than noise such as ventricle, DC components are generated in the first power spectrum and the second power spectrum. For this reason, if the first region and the second region are the same portion (same position), the normalization coefficient α is calculated by the same equation as the above equation (2). Further, in the first region and the second region, when the normalization coefficient α is calculated by the above equation (2) regardless of the presence or absence of the signal component at f = 0, P _base (0) and P _orig (0) are , F> 0, P _base (f) and P _orig (f) are respectively generated by extrapolation.

The frequency distribution generation unit 15 performs strong smoothing (smoothing) on the normalized first power spectrum Pn _orig (f) and the normalized second power spectrum Pn _base (f). This smoothing, frequency distribution generating unit 15 includes a first smoothing spectrum _{Pn_est orig} obtained by smoothing the first power spectrum _{Pn orig} normalized (f) (f), normalized second power spectrum _{Pn base} ( A second smoothed spectrum Pn_est _base (f) obtained by smoothing f) is generated. Note that the frequency distribution generation unit 15 performs the first smoothed spectrum and the second smoothed spectrum by curve approximation using a polynomial of a predetermined function, a Gaussian function, a Lithian function, or the like instead of smoothing. May be generated.

Further, the frequency distribution generation unit 15 may provide a cutoff frequency in the generation of the first smoothed spectrum Pn_est _orig (f) and the second smoothed spectrum Pn_est _base (f). When the first image (first interpolated image) and the second image (second interpolated image) are digital images, generally, high-frequency components are caused by aliasing distortion (distortion due to aliasing) due to a limited sampling frequency. ). For this reason, the frequency distribution generation unit 15 sets the cutoff frequency fc below the Nyquist frequency fn (fn ≧ fc) and cuts an excessive high frequency (f> fn ≧ fc).

The frequency distribution generation unit 15 usually adjusts the cutoff frequency to the lower sampling frequency side of the first image and the second image. Since the base image (second image) that is a past image often has a low sampling frequency, the frequency distribution generator 15 sets the first smoothed spectrum Pn_est _orig (f) to be equal to or lower than the sampling frequency fn of the second image. Determine the cutoff frequency at.

FIG. 4 shows a normalized first power spectrum Pn _orig (f) generated using, for example, the inverse Fourier transform of the noise profile shown in FIG. 3 corresponding to the first region and the second region, respectively. 2 is a diagram illustrating an example of the second power spectrum Pn _base (f), the first smoothed spectrum Pn_est _orig (f) obtained by smoothing the first and second power spectra, and the second smoothed spectrum Pn_est _base (f), respectively. . As shown in FIG. 4, the first smoothed spectrum Pn_est _orig (f) is a power spectrum obtained by smoothing the normalized first power spectrum Pn _orig (f). The second smoothed spectrum Pn_est _base (f) is a power spectrum obtained by smoothing the normalized second power spectrum Pn _base (f). Fc in FIG. 4 indicates a cut-off frequency.

The frequency distribution generation unit 15 generates the first frequency distribution S _orig (f) corresponding to the first image. Specifically, the frequency distribution generation unit 15 generates the first frequency distribution S _orig (f) by _performing inverse Fourier transform on the pixel value I _orig (x) of the first interpolation image or the first image. . That is, according to the above description, the frequency distribution generation unit 15 calculates the first frequency distribution S _orig (f) by the following expression, for example.

S _orig (f) = IFT [I _orig (x)]
The frequency distribution generator 15 outputs the first smoothed spectrum Pn_est _orig (f), the second smoothed spectrum Pn_est _base (f), and the first frequency distribution S _orig (f) to the correcting unit 17. The frequency distribution generation unit 15 _obtains the normalized first power spectrum Pn _orig (f), normalized second power spectrum Pn _base (f), and first frequency distribution S _orig (f). , It may be output to the correction unit 17. At this time, the frequency distribution generation unit 15 does not execute function approximation.

The correction unit 17 adjusts the noise level in the first region (for example, the standard deviation of noise) to the noise level in the second region based on the ratio between the approximate function Pn_est _orig (f) and Pn_est _base (f). The function PNR (f) is determined. Specifically, the correction unit 17 determines a conversion function PNR (f) that converts (matches or matches) the first smoothed spectrum Pn_est _orig (f) into the second smoothed spectrum Pn_est _base (f).

That is, the correction unit 17 determines the conversion function PNR (f) by dividing the second smoothed spectrum Pn_est _base (f) by the first smoothed spectrum Pn_est _orig (f). The conversion function PNR (f) corresponds to, for example, a noise power ratio (Power Noise Ratio). The conversion function PNR (f) is determined by the following equation. Note that the high frequency side of the conversion function PNR (f) is smoothly reduced below the cutoff frequency fc.
PNR (f) = Pn_est _base (f) / Pn_est _orig (f) (3)
In the above equation (3), it is necessary to devise so that division by zero does not occur. For example, in the frequency distribution generation unit 15 and FIG. 4, since the cutoff frequency fc is higher than the Nyquist frequency fn, Pn_est _orig (f) is non-zero.

In addition, when the function approximation is not performed on the normalized first power spectrum Pn _orig (f) and the normalized second power spectrum Pn _base (f), the correction unit 17 may, for example, In this way, the conversion function PNR (f) can be determined. The correction unit 17 uses the first power spectrum Pn normalized by an iterative operation (an operation such as a successive approximation) to which an algorithm (such as an optimization algorithm) for solving an optimization problem such as L1 or L2 norm optimization is applied. _The conversion function PNR (f) is determined while comparing _orig (f) and the normalized second power spectrum Pn _base (f).

FIG. 5 is a diagram illustrating an example of a conversion function for converting the first smooth spectrum Pn_est _orig (f) into the second smooth spectrum Pn_est _base (f). Fc shown in FIG. 5 is a cutoff frequency. In PNR (f) corresponding to FIG. 4, there is a straight line of PNR (f) = 1 between the value less than the maximum value of PNR (f) and the value of PNR (f) greater than α in FIG.

The correcting unit 17 corrects the first frequency distribution S _orig (f) using the conversion function PNR (f). The correction unit 17 outputs the corrected first frequency distribution S _cor (f) to the image generation unit 19. Specifically, the correction unit 17, by multiplying the conversion function PNR (f) to the first frequency distribution _{S orig} (f), corrects the first frequency distribution _{S orig} with (f). For example, the correction unit 17 corrects the first frequency distribution S _orig (f) according to the following equation.

S _cor (f) = S _orig (f) × PNR (f) (4)
FIG. 6 is a diagram illustrating an example of the power spectrum Pn _{orig_cor} (f) of the first frequency distribution corrected using the conversion function. Note that Pn _{orig_cor} (f) in FIG. 6 is normalized by the maximum value P _base (0) of the second power spectrum. Therefore, as shown in FIG. 6, the intersection of the vertical axis of the graph and Pn _{orig_cor} (f) is 1.

The image generator 19 applies the Fourier transform FT to the corrected first frequency distribution S _cor (f), thereby correcting the first image (first interpolated image) I _cor (x) (hereinafter referred to as “the first image”). , Referred to as a corrected first image). Specifically, the image generator 19 generates a corrected first image according to the following formula.

I _cor (x) = FT [S _cor (f)] (5)
The above formulas (1) to (5) are shown as one-dimensional formulas for the sake of simplicity, but the first image (or second image) to be corrected is a two-dimensional image or a three-dimensional image. Therefore, the above expressions (1) to (5) are expressions corresponding to the dimensions of the image. For example, when the correction target dimension is two-dimensional, the main correction is performed for each of the x and y directions. In addition, when the dimension to be corrected is three-dimensional, the main correction is performed for each of the x, y, and z directions, so that the higher dimension is performed as a more accurate correction.

  The image generation unit 19 outputs the generated corrected first image to the display unit 21. The image generation unit 19 may output the corrected first image to the storage unit 27. The image generation unit 19 generates an image (hereinafter referred to as a time-varying image) indicating a change with time of the corrected first image with respect to the second image based on the corrected first image and the second image. Also good. The image generation unit 19 outputs the temporally changing image to the display unit 21 and the storage unit 27.

  The display unit 21 displays various medical images generated by the image generation unit 19 and medical images stored in the storage unit 27. The display unit 21 displays various input screens related to the input operation.

  The interface unit 23 is an interface related to the network A, an external storage device (not shown), and the like. The interface unit 23 is connected to a medical image capturing apparatus and a medical image storage apparatus (not shown) via an electronic communication line, typically a local area network (hereinafter referred to as a LAN). . Volume data and image data generated by the medical image photographing apparatus, or volume data and image data stored by the medical image storage apparatus are stored in the storage unit 27 via the interface unit 23. Further, the medical image data and volume data corrected by the medical image processing apparatus 1 can be transferred to another device such as a medical image storage device (not shown) via a network.

  The input unit 25 inputs various instructions, commands, information, selections, settings, and the like from an operator or the like to the control unit 29. Specifically, the input unit 25 includes input devices such as a trackball, a switch button, a mouse, a mouse / wheel, and a keyboard for inputting the various instructions / commands / information / selection / settings. The input device may be a touch panel that covers the display screen in the display unit 21. The input unit 25 inputs a selection operation between the first image to be corrected and the second image collected by the same kind of modality in the past from the first image via the input device. The input unit 25 can also input the positions, shapes, and sizes of the first area and the second area.

  The storage unit 27 includes a first image and a second image transferred via the interface unit 23, various frequency distributions generated by the frequency distribution generation unit 15, a power spectrum, a conversion function determined by the correction unit 17, and an image generation Various medical images generated by the unit 19 are stored. The storage unit 27 may store the shapes and sizes of the first region and the second region used in the region setting unit 13. The storage unit 27 stores various control programs used by the control unit 29. Further, the storage unit 27 is a correction program related to a frequency distribution correction process to be described later, for example, interpolation processing, setting of first and second regions, generation of frequency distribution and various power spectra, normalization and smoothing of power spectrum, conversion At least one program relating to function determination, frequency distribution correction, generation of corrected image, and the like is stored.

  The control unit 29 includes a central processing unit (Central Processing Unit: hereinafter referred to as CPU), a memory, and the like which are not illustrated. The control unit 29 functions as the center of the medical image processing apparatus 1. Specifically, the control unit 29 reads various programs from the storage unit 27 in accordance with an input operation input via the input unit 25 and expands the program on its own memory. The control unit 29 controls each unit of the medical image processing apparatus 1 by executing the developed program. Further, for example, the control unit 29 reads out and executes the correction program from the storage unit 27, thereby executing each unit such as the interpolation processing unit 11, the region setting unit 13, the frequency distribution generation unit 15, the correction unit 17, and the image generation unit 19. To control.

  The medical image processing apparatus 1 includes various modalities, for example, a magnetic resonance imaging (MRI) apparatus, an X-ray computed tomography (CT) apparatus, a single photon emission computed tomography (Single Photon Emission Imaging). It may be mounted on a console (such as a console) such as a computerized tomography (SPECT) apparatus, a positron emission computed tomography (PET) apparatus, or an X-ray diagnostic apparatus. The medical image processing apparatus 1 can also be realized as a workstation. Further, the medical image processing apparatus 1 may be realized as, for example, a thin client and a server in a cloud system.

(Frequency distribution correction function)
The frequency distribution correction function is a frequency distribution of a medical image to be corrected so that the noise level of the medical image (first image) to be corrected matches the noise level of the medical image (second image) to be a reference. It is a function to correct. Hereinafter, processing related to the frequency distribution correction function (hereinafter referred to as frequency distribution correction processing) will be described.

FIG. 7 is a flowchart illustrating an example of a procedure related to frequency distribution correction processing.
In response to an operator instruction via the input unit 25, the selected first image and second image are read from the storage unit 27. Note that the selected first image and second image may be read from a medical image storage device, various modalities, etc. (not shown) via the interface unit 23 and the network.

  A spatially uniform first region is set based on a plurality of pixel values in the read first image (step Sa1). A spatially uniform second region is set based on the plurality of pixel values in the read second image (step Sa2).

  Based on the plurality of pixel values in the first region, a first power spectrum corresponding to the first region frequency distribution is generated (step Sa3). Based on the plurality of pixel values in the second region, a second power spectrum corresponding to the second region frequency distribution is generated (step Sa3). Each of the first and second power spectra is normalized using the maximum value of the second power spectrum (step Sa4).

  A first smoothed spectrum is generated by smoothing the normalized first power spectrum (step Sa5). By smoothing the normalized second power spectrum, a second smooth spectrum is generated (step Sa5). A conversion function for converting the first smooth spectrum into the second smooth spectrum is determined (step Sa6).

  A first frequency distribution is generated based on a plurality of pixel values in the entire area of the first image (first interpolation image) (step Sa7). The first frequency distribution is corrected using the conversion function (step Sa8). Based on the corrected first frequency distribution, a corrected first image obtained by correcting the first image is generated (step Sa9).

(First modification)
The difference from the first embodiment is that the first power spectrum is used as a reference.
The frequency distribution generation unit 15 normalizes the first power spectrum by dividing the first power spectrum P _orig (f) by the maximum value P _orig (0) of the first power spectrum. In addition, the frequency distribution generation unit 15 normalizes the second power spectrum by dividing the second power spectrum P _base (f) by the maximum value P _orig (0) of the first power spectrum.

That is, the normalized first power spectrum Pn _orig (f) is calculated as P _orig (f) / P _orig (0). The normalized second power spectrum Pn _base (f) is calculated as P _base (f) / P _orig (0). The frequency distribution generator 15 calculates a normalization coefficient α. For example, the normalization coefficient α is calculated as in the following equation.

α = P _orig (0) / P _base (0)
The frequency distribution generator 15 generates a second frequency distribution S _base (f) corresponding to the second image. Specifically, the frequency distribution generation unit 15 generates the second frequency distribution S _base (f) by performing inverse Fourier transform on the pixel value I _base (x) of the second interpolation image or the second image. . That is, according to the above description, the frequency distribution generation unit 15 calculates the second frequency distribution S _base (f) by the following expression, for example.

S _base (f) = IFT [I _base (x)]
The correction unit 17 determines a function PNR (f) for adjusting the noise level in the second region to the noise level in the first region, based on the ratio between the approximate function Pn_est _orig (f) and Pn_est _base (f). . Specifically, the correction unit 17 determines a conversion function PNR (f) that converts (matches or matches) the second smoothed spectrum Pn_est _base (f) into the first smoothed spectrum Pn_est _orig (f).

That is, the correction unit 17 determines the conversion function PNR (f) by dividing the first smoothed spectrum Pn_est _orig (f) by the second smoothed spectrum Pn_est _base (f). The conversion function PNR (f) corresponds to, for example, a noise power ratio (Power Noise Ratio). The conversion function PNR (f) is determined by the following equation.
PNR (f) = Pn_est _orig (f) / Pn_est _base (f)
In the above equation, for example, the frequency distribution generation unit 15 may determine the cutoff frequency fc so that division by zero does not occur. It should be noted that the conversion function is interpolated if various information such as the frequency distribution of interpolation at the time of generation of the first image and the second image, the number of interpolations, the interpolation algorithm, the interpolation size, the type and shape of the applied filter are known. The conversion function may be determined using the frequency distribution, the number of interpolations, an adapted filter, and the like.

The correcting unit 17 corrects the second frequency distribution S _base (f) using the conversion function PNR (f). The correction unit 17 outputs the corrected second frequency distribution S _cor2 (f) to the image generation unit 19. Specifically, the correction unit 17, by multiplying the conversion function PNR (f) in the second frequency distribution _{S base} (f), corrects the second frequency distribution _{S base} (f). For example, the correction unit 17 corrects the second frequency distribution S _base (f) according to the following equation.

S _cor2 (f) = S _base (f) × PNR (f)
The image generation unit 19 applies the Fourier transform FT to the corrected second frequency distribution S _cor2 (f), thereby correcting an image I _cor2 (x) (hereinafter _referred to as a second image (second interpolation image)). , Referred to as a corrected second image). Specifically, the image generation unit 19 generates a corrected second image according to the following equation.

I _cor2 (x) = FT [S _cor2 (f)]
The image generation unit 19 outputs the generated corrected second image to the display unit 21. The image generation unit 19 may output the corrected second image to the storage unit 27. Note that the image generation unit 19 may generate an image indicating a change with time of the corrected second image with respect to the first image based on the corrected second image and the first image.

(Second modification)
The difference between the first embodiment and the first modification is that a power spectrum corresponding to a predetermined frequency distribution is used as a reference.
The storage unit 27 stores a power spectrum (hereinafter referred to as a predetermined power spectrum) P _predet (f) corresponding to a predetermined frequency distribution. The predetermined frequency distribution is, for example, a frequency distribution corresponding to white noise or a frequency distribution corresponding to fixed pattern noise. In addition, the storage unit 27 stores a normalized power spectrum (hereinafter, normalized power spectrum Pn _predet (f). The stored power spectrum P _predet (f) is stored in the frequency distribution generating unit 15. _Is read from the storage unit 27 to the frequency distribution generation unit 15. The stored normalized power spectrum Pn _predet (f) is read from the storage unit 27 to the correction unit 17 by the correction unit 17.

The frequency distribution generator 15 generates the first frequency distribution S _orig (f) and the second frequency distribution S _base (f). The frequency distribution generator 15 normalizes the first power spectrum by dividing the first power spectrum P _orig (f) by the maximum value P _predet (0) of the predetermined power spectrum. The frequency distribution generation unit 15 normalizes the second power spectrum by dividing the second power spectrum P _base (f) by the maximum value P _predet (0) of the predetermined power spectrum. The normalized first power spectrum and the normalized second power spectrum are output to the correction unit 17.

Specifically, the normalized first power spectrum Pn _orig (f) is calculated as P _orig (f) / P _predet (0). The normalized second power spectrum Pn _base (f) is calculated as P _base (f) / P _predet (0). The frequency distribution generation unit 15 calculates a normalization coefficient α1 corresponding to the first power spectrum. The frequency distribution generation unit 15 calculates a normalization coefficient α2 corresponding to the second power spectrum. For example, the normalization coefficient α1 is calculated as in the following equation.

α1 = P _predet (0) / P _orig (0)
Further, the normalization coefficient α2 is calculated as in the following formula, for example.

α2 = P _predet (0) / P _base (0)
The correction unit 17 determines the noise level in the first region based on the ratio between the normalized first power spectrum Pn_est _orig (f) and the normalized power spectrum Pn _predet (f) , Referred to as a predetermined noise level) to determine a first conversion function PNR1 (f). The correcting unit 17 adjusts the noise level in the second region to a predetermined noise level based on the ratio between the normalized second power spectrum Pn_est _base (f) and the normalized power spectrum Pn _predet (f). A two-transform function PNR2 (f) is determined.

Specifically, the correction unit 17 determines a first conversion function PNR (f) that converts (matches or matches) the first smoothed spectrum Pn_est _orig (f) into the normalized power spectrum Pn _predet (f). That is, the correction unit 17 determines the first conversion function PNR1 (f) by dividing the normalized power spectrum Pn _predet (f) by the first smoothed spectrum Pn_est _orig (f). The correction unit 17 determines a second conversion function PNR (f) that converts (matches or matches) the second smoothed spectrum Pn_est _base (f) into the normalized power spectrum Pn _predet (f). That is, the correction unit 17 determines the second conversion function PNR (f) by dividing the normalized power spectrum Pn _predet (f) by the second smoothed spectrum Pn_est _base (f).

  The first conversion function PNR1 (f) and the second conversion function PNR2 (f) correspond to, for example, a noise power ratio (Power Noise Ratio). The first conversion function PNR1 (f) and the second conversion function PNR2 (f) are respectively determined by the following two equations.

PNR1 (f) = Pn _predet (f) / Pn_est _orig (f)
PNR2 (f) = Pn _predet (f) / Pn_est _base (f)
The correcting unit 17 corrects the first frequency distribution S _orig (f) using the first conversion function PNR1 (f). The correcting unit 17 corrects the second frequency distribution S _base (f) using the second conversion function PNR2 (f). The correction unit 17 outputs the corrected first frequency distribution S _cor1 (f) and the corrected second frequency distribution S _cor2 (f) to the image generation unit 19.

Specifically, the correction unit 17, by multiplying the first transformation function PNR1 (f) to the first frequency distribution _{S orig} (f), corrects the first frequency distribution _{S orig} with (f). For example, the correction unit 17 corrects the first frequency distribution S _orig (f) according to the following equation.

S _cor1 (f) = S _orig (f) × PNR1 (f)
The correction unit 17, by multiplying the second transformation function PNR2 (f) in the second frequency distribution _{S base} (f), corrects the second frequency distribution _{S base} (f). For example, the correction unit 17 corrects the second frequency distribution S _base (f) according to the following equation.

S _cor2 (f) = S _base (f) × PNR2 (f)
The image generation unit 19 applies the Fourier transform FT to the corrected first frequency distribution S _cor1 (f) to correct the corrected first image I _cor1 (x ). Specifically, the image generator 19 generates a corrected first image according to the following formula.

I _cor1 (x) = FT [S _cor1 (f)]
The image generation unit 19 applies the Fourier transform FT to the corrected second frequency distribution S _cor2 (f), thereby correcting the second image (second interpolated image) to _obtain a corrected second image I _cor2 (x ). Specifically, the image generation unit 19 generates a corrected second image according to the following equation.

I _cor2 (x) = FT [S _cor2 (f)]
The image generation unit 19 outputs the corrected first image and the corrected second image to the display unit 21. The image generation unit 19 may output the corrected first image and the corrected second image to the storage unit 27. Note that the image generation unit 19 may generate an image indicating a temporal change of the corrected first image with respect to the corrected second image based on the corrected first image and the corrected second image.
According to the configuration described above, the following effects can be obtained.
According to the medical image processing apparatus 1 according to the present embodiment, the first area and the second area, which are unstructured and spatially uniform areas, are set as the first image and the second image, respectively. The corrected first image in which the noise level of the first image is matched with the noise level of the second image can be generated based on the first power spectrum and the second power spectrum in the set first region. In particular, the frequency distribution correction processing according to the present embodiment is effective when the image reconstruction method and the filter are linear algorithms.

  Moreover, according to the 1st modification which concerns on this embodiment, the correction | amendment 2nd image which match | combined the noise level of the 2nd image with the noise level of the 1st image based on the 1st power spectrum and the 2nd power spectrum. Can be generated.

  Further, according to the second modification example of the present embodiment, the noise level of the first image and the first power spectrum, the second power spectrum, and the power spectrum related to the predetermined frequency distribution are set to a predetermined noise level. It is possible to generate a corrected first image and a corrected second image that are respectively combined with the noise levels of the two images.

  In other words, according to the medical image processing apparatus 1, a corrected image excluding the influence of changes in the image derived from the apparatus such as the collection apparatus (modality) of the first image and the second image, and the parameters of the collection conditions is excluded from the image. It can occur even if there is no information. This makes it possible to display the anatomical / functional change of the subject over time, that is, the essential change over time, without intentionally blurring the medical image.

  From the above, according to the medical image processing apparatus 1, including a field such as VBM, when tracking changes in images over time, temporal changes due to factors inherent to the patient are excluded by excluding factors derived from the imaging device. Can be tracked. Accordingly, the operator can track the original temporal change in more detail in the medical image related to the subject.

(Second Embodiment)
The difference from the first embodiment is that the first region and the second region are regions that change little over time or regions that do not change over time (hereinafter referred to as time-invariant regions), and other than noise. It has a signal component.
FIG. 8 is a configuration diagram showing the configuration of the medical image processing apparatus 1 according to the present embodiment. The difference from FIG. 1 resides in having an edge detection unit 10.

  The edge detection unit 10 detects a plurality of first edges in the first image (or the first interpolation image) by a predetermined edge detection process. The edge detection unit 10 detects a plurality of second edges in the second image (or second interpolation image) by a predetermined edge detection process.

  The storage unit 27 stores a predetermined threshold. The predetermined threshold is a pixel value corresponding to an edge that defines a temporally unchanged region in the first image and the second image. The time-invariant region is a region having a substantially constant (flat) pixel value sandwiched between two edges, and indicates, for example, a predetermined phantom imaged together with the subject in the first image and the second image. A region, a region indicating a bone shape (for example, a region corresponding to a skull in a brain image), a region mainly indicating water (for example, a CSF in a brain image), and the like. Hereinafter, in order to simplify the description, it is assumed that the time-invariant region is a region corresponding to the phantom or a region corresponding to the CSF.

  The region setting unit 13 sets a pixel value range in which two opposite sides in the first region are orthogonal to the first edge and sandwiched by a predetermined threshold as the first region in the first interpolation image. The region setting unit 13 sets, in the second interpolation image, a pixel value range in which two opposite sides in the second region are orthogonal to the second edge and sandwiched by a predetermined threshold as the second region. The first region and the second region are preferably regions having the same pixel value range (for example, phantom, bone, CSF, water, etc.). When phantoms are used as the first region and the second region, the phantom images are collected simultaneously with the collection of the first image and the second image or at the smallest possible time difference in order to perform stable frequency distribution correction. Is preferred. Note that the first region and the second region may be set by bone, water extraction, or the like that changes little over time. Further, the first area and the second area may be set by an instruction from the operator via the input unit 25.

  FIG. 9 is a diagram illustrating an example of setting the first region or the second region in the first interpolation image and the second interpolation image. In FIG. 9, f indicates a phantom. A line segment dd ′ in FIG. 9 shows an example of the first area or the second area set for the phantom f. A line segment ee ′ in FIG. 9 shows an example of the first region or the second region set for the region corresponding to the skull and the region corresponding to the CSF.

  FIG. 10 is a diagram illustrating an example of a change in pixel value (actual signal profile) due to a plurality of pixel values respectively corresponding to a plurality of pixels arranged one-dimensionally on the line segment ee ′ in FIG. 9. The horizontal axis in FIG. 10 indicates the pixel position x. The vertical axis in FIG. 10 indicates the pixel value I (x) at the pixel position x in the image I. In FIG. 10, the first region and the second region are set as regions sandwiched by a line segment ff ′, for example.

FIG. 11 is a diagram illustrating an example of a change (time-invariant profile) of a pixel value on the line segment dd ′ in FIG. 9 or the extracted portion (ff ′) in FIG. The horizontal axis in FIG. 11 indicates the pixel position x. The vertical axis in FIG. 11 indicates the pixel value I (x) at the pixel position x in the image I. In FIG. 11, I _orig (x) indicates a profile of a plurality of pixel values respectively corresponding to a plurality of pixels located in the line segment dd ′ or the line segment ff ′ in the first image. In FIG. 11, I _base (x) indicates a profile of a plurality of pixel values respectively corresponding to a plurality of pixels located in the line segment dd ′ or the line segment ff ′ in the second image. 9 to 11, the first region and the second region are described as one-dimensional profiles, but may be two-dimensional or three-dimensional profiles.

The frequency distribution generator 15 generates the absolute value | S1 _orig (f) | of the first region frequency distribution S1 _orig (f) corresponding to the first region. Specifically, the frequency distribution generation unit 15 performs an inverse Fourier transform on the plurality of pixel values I1 _orig (x) in the first region to obtain an absolute value, thereby obtaining the absolute value of the first region frequency distribution. | S1 _orig (f) | That is, according to the above description, the frequency distribution generation unit 15 calculates the absolute value | S1 _orig (f) | of the first region frequency distribution by the following expression, for example.

| S1 _orig (f) | = | IFT [I1 _orig (x)] |
The frequency distribution generation unit 15 generates the absolute value | S2 _base (f) | of the second region frequency distribution S2 _base (f) corresponding to the second region. Specifically, the frequency distribution generation unit 15 performs an inverse Fourier transform on the plurality of pixel values I2 _base (x) in the second region to obtain an absolute value, thereby obtaining the absolute value of the second region frequency distribution. | S2 _base (f) | is generated. That is, according to the above description, the frequency distribution generation unit 15 calculates | S2 _base (f) | of the absolute value of the second region frequency distribution, for example, using the following equation.

| S2 _base (f) | = | IFT [I2 _base (x)] |
The frequency distribution generation unit 15 divides the absolute value of the first region frequency distribution by the absolute value | S2 _base (0) | of the second region frequency distribution by dividing the absolute value of the first region frequency distribution by the maximum value | S2 _base (0) | Normalize. Further, the frequency distribution generation unit 15 divides the absolute value of the second region frequency distribution by the maximum absolute value | S2 _base (0) | of the second region frequency distribution to thereby obtain the absolute value of the second region frequency distribution. Is normalized.

That is, the normalized absolute value of the first region frequency distribution (hereinafter, referred to as a first normalized frequency _{distribution) | Sn1 orig (f) | is, | S1 orig (f) |} / | S2 base (0) | Is calculated as The absolute value of the normalized second region frequency distribution (hereinafter referred to as the second normalized frequency distribution) | Sn2 _base (f) | is calculated as | S2 _base (f) | / | S2 _base (0) | Is done. The frequency distribution generator 15 calculates a normalization coefficient α. For example, the normalization coefficient α is calculated as in the following equation.
α = | S2 _base (0) | / | S1 _orig (0) |
The frequency distribution generating unit 15 outputs the first normalized frequency distribution | Sn1 _orig (f) | and the second normalized frequency distribution | Sn2 _base (f) | to the correcting unit 17.

FIG. 12 shows the relationship between the first normalized frequency distribution | Sn1 _orig (f) | and the second normalized frequency distribution | Sn2 _base (f) | generated by normalizing the absolute value of the inverse Fourier transform of the time-invariant profile. It is a figure which shows an example. In the first normalized frequency distribution | Sn1 _orig (f) | and the second normalized frequency distribution | Sn2 _base (f) | in FIG. 12, the influence of noise is relatively small compared to the first embodiment. The cost of smoothing and function approximation in 1 is smaller than that in the first embodiment. For the sake of simplicity, smoothing or function approximation is not mentioned in the present embodiment. However, smoothing or function approximation is appropriately performed for | S1 _orig (f) | and | S2 _base (f) |. May be executed.

Correcting unit 17, first normalized frequency distribution _{| Sn1 orig} (f) _| and the second normalized frequency distribution _{| Sn2 base} (f) _| by the ratio of the noise level of noise level in the first region in the second region A function F _cor (f) for adjusting to is determined. Specifically, the correction unit 17, first normalized frequency distribution _{| Sn1 orig} (f) _|, and into a second smoothing spectrum _{Pn_est base} (f) (matched or matched to) the conversion function _F cor (f) To decide.

That is, the correction unit 17 determines the conversion function F _cor (f) by dividing the second normalized frequency distribution | Sn2 _base (f) | by the first normalized frequency distribution | Sn1 _orig (f) |. . Specifically, the conversion function F _cor (f) is determined by the following equation. Note that the high frequency side of the conversion function F _cor (f) is smoothly reduced below the cutoff frequency fc.

F _cor (f) = | Sn2 _base (f) | / | Sn1 _orig (f) |
In the above equation, it is necessary to devise so that division by zero does not occur. For example, in the frequency distribution generation unit 15 and FIGS. 12 to 13, since the cutoff frequency fc is higher than the Nyquist frequency fn, | Sn1 _orig (f) | is non-zero.

FIG. 13 is a diagram illustrating an example of the conversion function F _cor (f) for converting the first normalized frequency distribution into the second normalized frequency distribution. Fc shown in FIG. 13 is a cutoff frequency.

The correcting unit 17 corrects the first frequency distribution S _orig (f) using the conversion function F _cor (f). The correction unit 17 outputs the corrected first frequency distribution S _cor (f) to the image generation unit 19. Specifically, the correction unit 17, by multiplying the conversion function _{F cor} (f) to the first frequency distribution _{S orig} (f), corrects the first frequency distribution _{S orig} with (f). For example, the correction unit 17 corrects the first frequency distribution S _orig (f) according to the following equation.

S _cor (f) = S _orig (f) × F _cor (f)
FIG. 14 is a diagram illustrating an example of the absolute value | S _cor (f) | of the first frequency distribution corrected using the conversion function F _cor (f). Fc shown in FIG. 14 is a cutoff frequency.

(Frequency distribution correction function)
FIG. 15 is a flowchart illustrating an example of a procedure related to frequency distribution correction processing.
A plurality of first edges are detected in the first image (step Sb1). A plurality of second edges are detected in the second image (step Sb2). A region where the opposite side of the first region is orthogonal to the first edge and the change with time is small is set as the first region in the first image (step Sb3). A region where the opposite side of the second region is orthogonal to the second edge and changes little over time is set as the second region in the second image (step Sb3).

  Based on the plurality of pixel values in the first region, an absolute value of the first region frequency distribution is generated (step Sb4). Based on the plurality of pixel values in the second region, an absolute value of the second region frequency distribution is generated (step Sb4). Using the maximum absolute value of the second region frequency distribution, each absolute value of the first and second region frequency distributions is normalized (step Sb5).

  A conversion function for converting the first normalized frequency distribution into the second normalized frequency distribution is determined (step Sb7). A first frequency distribution is generated based on a plurality of pixel values in the entire area of the first image (step Sb8). The first frequency distribution is corrected using the conversion function (step Sb9). Based on the corrected first frequency distribution, a corrected first image obtained by correcting the first image is generated (step Sb10).

According to the configuration described above, the following effects can be obtained.
According to the medical image processing apparatus 1 according to the present embodiment, the first area and the second area, which are unstructured areas that do not change with time, are set as the first image and the second image, respectively. Based on the first region frequency distribution in the first region and the second region frequency distribution in the second region, a corrected first image in which the noise level of the first image is matched with the noise level of the second image can be generated. .

  In the frequency distribution correction process according to the present embodiment, when it is difficult to estimate the noise characteristics and the image structure characteristics, specifically, for example, a portion where the pixel value is flat except for a complicated structure portion. This is more effective than the frequency distribution correction processing according to the first embodiment in the case where a non-linear algorithm that strongly smoothes only is used.

  The frequency distribution correction process according to the first embodiment and the frequency distribution correction process according to the second embodiment can be used in combination. For example, normally, when a base image (second image) acquired in the past with respect to a target image (first image) is generated by a conventional medical image generation device, image generation is performed using a linear algorithm. It is likely that it is being executed. In addition, since an image by a recent apparatus that can be a target image often masks an air portion, it is difficult to obtain a power spectrum by the frequency distribution correction processing according to the first embodiment. In such a case, the power distribution is acquired from the base image (second image) using the frequency distribution correction process according to the first embodiment, and the target is obtained using the frequency distribution correction process according to the second embodiment. The first frequency distribution can be corrected by acquiring the power spectrum from the image (first image).

  From the above, according to the medical image processing apparatus 1 according to the first and second embodiments, in order to exclude apparatus-dependent fluctuations when comparing temporal changes with past images, that is, in temporal changes. In order to exclude factors derived from the imaging device, the frequency distribution can be corrected. Thereby, according to this medical image processing apparatus 1, the operator can track the anatomical temporal change by the patient's original factor with high accuracy.

(Third embodiment)
The difference between the first embodiment and the second embodiment is that the filter function related to the power spectrum characteristics of the first image and the second image, the interpolation function related to the interpolation processing, and various information related to the number of interpolations are known. For example, using this various information, a correction function such as the conversion function F _cor (f), the first conversion function PNR1 (f), or the second conversion function PNR2 (f) used in the correction unit 17 is generated. It is to correct.

  The storage unit 27 stores various information related to the filter function related to the power spectrum characteristics of the first image and the second image, the interpolation function related to the interpolation processing, and the number of interpolations.

The correction unit 17 reads the various types of information from the storage unit 27, and based on the various types of information, the conversion function F _cor (f), the first conversion function PNR1 (f), the second conversion function PNR2 (f), or the like. Determine the correction function. For example, the correction unit 17 determines the correction function based on the filter function, the interpolation function, the number of interpolations, and the like regarding the first region frequency distribution and the second region frequency distribution. Next, the correction unit 17 corrects at least one of the first frequency distribution and the second frequency distribution using the determined correction function.

  From the above, the medical image processing apparatus 1 according to the third embodiment further has the following effects in addition to the effects according to the first and second embodiments. Compared to the first and second embodiments, the medical image processing apparatus 1 according to the third embodiment performs various calculations (power spectrum characteristic measurement) related to various power spectrum characteristics in the frequency distribution generator 15 and the like. It becomes unnecessary, and the frequency distribution, power spectrum, etc. can be corrected with higher accuracy.

  In addition, each function according to the embodiment can also be realized by installing a program for executing the processing in a computer such as a workstation and developing the program on a memory. At this time, a program capable of causing the computer to execute the method is stored in a storage medium such as a magnetic disk (floppy (registered trademark) disk, hard disk, etc.), an optical disk (CD-ROM, DVD, etc.), or a semiconductor memory. It can also be distributed.

  As mentioned above, although some embodiment of this invention was described, these embodiment is shown as an example and is not intending limiting the range of invention. These novel embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, and are included in the invention described in the claims and the equivalents thereof.

  DESCRIPTION OF SYMBOLS 1 ... Medical image processing apparatus, 10 ... Edge detection part, 11 ... Interpolation processing part, 13 ... Area setting part, 15 ... Frequency distribution generation part, 17 ... Correction part, 19 ... Image generation part, 21 ... Display part, 23 ... Interface unit, 25... Input unit, 27... Storage unit, 29.

Claims (10)

An area setting unit that sets a first area corresponding to a part of the first image and a second area corresponding to a part of the second image collected in the past from the first image;
A frequency distribution generator that generates a first region frequency distribution based on a plurality of pixel values in the first region and generates a second region frequency distribution based on a plurality of pixel values in the second region;
At least one of a first frequency distribution corresponding to the first image and a second frequency distribution corresponding to the second image with reference to the first region frequency distribution, the second region frequency distribution, or a predetermined frequency distribution. A correction unit for correcting one;
A medical image processing apparatus comprising: The region setting unit
A spatially uniform region in the first image or a region that does not change over time is set as the first region;
Setting a spatially uniform area in the second image or an area that does not change over time as the second area;
The medical image processing apparatus according to claim 1. The region setting unit
Based on the pixel value of the first image, the first area is set,
Setting the second region based on a pixel value of the second image;
The medical image processing apparatus according to claim 1 or 2. The frequency distribution generator is
Generating a first smoothed spectrum by smoothing a power spectrum corresponding to the first region frequency distribution;
Generating a second smoothed spectrum by smoothing a power spectrum corresponding to the second region frequency distribution;
The correction unit is
Determining a conversion function for converting the first smoothed spectrum to the second smoothed spectrum;
Correcting the first frequency distribution using the conversion function;
The medical image processing apparatus according to claim 1, wherein the medical image processing apparatus is a medical image processing apparatus. The frequency distribution generator is
Generating a first smoothed spectrum by smoothing a power spectrum corresponding to the first region frequency distribution;
Generating a second smoothed spectrum by smoothing a power spectrum corresponding to the second region frequency distribution;
The correction unit is
Determining a conversion function for converting the second smooth spectrum to the first smooth spectrum;
Correcting the second frequency distribution using the conversion function;
The medical image processing apparatus according to claim 1, wherein the medical image processing apparatus is a medical image processing apparatus. The frequency distribution generator is
Generating a first smoothed spectrum by smoothing a power spectrum corresponding to the first region frequency distribution;
Generating a second smoothed spectrum by smoothing a power spectrum corresponding to the second region frequency distribution;
The correction unit is
Determining a first conversion function for converting the first smooth spectrum into a power spectrum corresponding to the predetermined frequency distribution;
Determining a second conversion function for converting the second smooth spectrum into a power spectrum corresponding to the predetermined frequency distribution;
Correcting the first frequency distribution using the first transformation function;
Correcting the second frequency distribution using the second transformation function;
The medical image processing apparatus according to claim 1, wherein the medical image processing apparatus is a medical image processing apparatus. The frequency distribution generator is
Using the maximum value of the first power spectrum corresponding to the first domain frequency distribution and the maximum value of the second power spectrum corresponding to the second domain frequency distribution, the first power spectrum and the second power spectrum, Normalize
The correction unit is
Correcting at least one of the first frequency distribution and the second frequency distribution based on the first power spectrum, the second power spectrum, or a power spectrum corresponding to the predetermined frequency distribution;
The medical image processing apparatus according to any one of claims 1 to 6. An edge detector for detecting a plurality of first edges in the first image and detecting a plurality of second edges in the second image;
The region setting unit
A region corresponding to a pixel value range in which two opposing sides in the first region are orthogonal to the first edge and defined by a predetermined threshold is set as the first region;
Setting an area in which the two sides in the second area are orthogonal to the second edge and corresponding to the pixel value range as the second area;
The medical image processing apparatus according to claim 1, wherein: The correction unit is
A correction function is determined based on a filter function, an interpolation function, and the number of interpolations related to the first region frequency distribution and the second region frequency distribution, and the first frequency distribution and the second frequency are determined using the determined correction function. Correcting at least one of the distribution,
The medical image processing apparatus according to claim 1. Setting a first region corresponding to a part of the first image and a second region corresponding to a part of the second image collected in the past from the first image;
Generating a first region frequency distribution based on a plurality of pixel values in the first region;
Generating a second region frequency distribution based on a plurality of pixel values in the second region;
At least one of a first frequency distribution corresponding to the first image and a second frequency distribution corresponding to the second image with reference to the first region frequency distribution, the second region frequency distribution, or a predetermined frequency distribution. Correcting one,
A medical image processing method comprising: