JP2008067916A - Medical image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a medical image processor which prepares a long length image or a difference image efficiently by automatically performing alignment with high precision. <P>SOLUTION: According to the correlation between a template area and a search area set to two images photographed while moving their photographing positions in the direction of the body axis of a patient so as to superpose partially, the two images are aligned roughly. Thereafter, within a processing area extracted from a part registered roughly, the two images are aligned locally according to a shape featured value concerning the continuity of the shape of the back bone. Thus, the images are aligned automatically and precisely. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a medical image processing apparatus, and more particularly to a medical image processing apparatus that creates a new image by joining and subtracting two or more images after aligning them.

  When the subject is imaged twice in the body axis direction and the two images are overlapped and joined together to create a long image, the difference between the two images taken at the same site and taken at different times is taken to obtain a normal structure For example, when creating a difference image from which changes over time are extracted by deleting the image, it is necessary to accurately align two photographs that are the basis for creating a long image, a difference image, and the like.

  However, when an operator who observes an image visually aligns the positions of the two images, skill is required for the alignment, so in the case of an unfamiliar operator, the alignment accuracy is low, There is a possibility that an accurate long image or difference image cannot be created. Moreover, since alignment is a manual operation, there is a problem that it takes time.

In order to deal with this, Patent Document 1 discloses that the subject is moved a plurality of times in the body axis direction while the subject is in close contact with the detector, and the subject is imaged a plurality of times and the images are combined. A technique for creating a long image is disclosed.
JP 2005-270277 A

  However, Patent Document 1 has the following drawbacks. In Patent Document 1, since it is assumed that the subject does not move, there is a problem that when the subject moves, a long image whose position is shifted by the amount the subject moves is created. In addition, even when the subject does not move, there is a problem that an error occurs due to the movement of the detector in the body axis direction, and a long image whose position is shifted by that amount is generated. Further, when correcting the misalignment, there is a problem that the operator must perform the correction while confirming the image, and the efficiency is lowered.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a medical image processing apparatus that efficiently creates a long image or a difference image by automatically performing high-precision alignment. To do.

  In order to solve the above problem, the medical image processing apparatus according to claim 1 includes a first image obtained by imaging a subject, and a second image in which a part of the first image overlaps with the first image. Image acquisition means for acquiring the first attention area on the first image and the second region used for the alignment from the overlapping image when the first image and the second image are aligned First attention area extraction means for extracting a second attention area on the image, an image in the first attention area and an image in the second attention area extracted by the first attention area extraction means. And a first alignment means for aligning the first image and the second image based on the correlation, and a first image aligned by the first alignment means A predetermined processing area including the image of the alignment portion from the second image and the second image Processing area extracting means for extracting, and second positioning means for aligning the first image and the second image based on the shape feature quantity or density feature quantity of the image in the extracted processing area And image creating means for creating a new image by processing the first image and the second image based on the first image and the second image aligned by the second positioning means And display means for displaying the new image created by the image creating means.

  According to the medical image processing apparatus of claim 1, first, the first image and the second image that are imaged by moving the imaging position in the body axis direction of the subject so as to partially overlap each other. A first region of interest and a second region of interest are set for each, and based on their correlation, the first image and the second image are roughly aligned. Thereafter, local alignment between the first image and the second image is performed based on the shape feature amount or the density feature amount in the processing region extracted from the portion where the rough alignment has been performed.

  Thereby, highly accurate alignment can be performed automatically.

  According to a second aspect of the present invention, in the medical image processing apparatus according to the first aspect, the shape feature amount is a value obtained from continuity of the shape of the spine.

  According to the medical image processing apparatus of the second aspect, the first image and the second image respectively captured by moving the imaging position in the body axis direction of the subject so as to partially overlap each other. One attention area and a second attention area are set, and the first image and the second image are roughly aligned based on the correlation between them. Thereafter, local alignment between the first image and the second image is performed based on the shape feature amount related to the continuity of the shape of the spine in the processing region extracted from the portion where the rough alignment has been performed. Do.

  Thus, a long image can be easily and efficiently created by automatically performing high-precision positioning. In addition, the shape feature value related to the continuity of the shape of the spine, that is, the feature that the length of the spine constituting the spine is the same or longer as it goes from the head to the foot direction is used, so that the alignment in the body axis direction is precise. Can be done.

  The medical image processing apparatus according to claim 3 is the medical image processing apparatus according to claim 1, wherein the processing region extraction unit includes a first image and a second image that are aligned by the first alignment unit. The third attention area on the first image that is narrower than the first attention area and the fourth on the second image that is narrower than the second attention area at a plurality of positions in the alignment area with the image of FIG. And the second alignment means extracts the image in the extracted third region of interest obtained for each of the plurality of locations as the density feature amount and the extracted fourth region. The first image and the second image are aligned based on a value related to a correlation with an image in the region of interest.

  According to the medical image processing apparatus of the third aspect, the first image and the second image respectively obtained by moving the imaging position in the body axis direction of the subject so as to partially overlap each other. One attention area and a second attention area are set, and the first image and the second image are roughly aligned based on the correlation between them. Thereafter, a correlation between a plurality of third regions of interest on the first image and a plurality of fourth regions of interest on the second image extracted from the roughly aligned portion is obtained. Based on this, local alignment between the first image and the second image is performed.

  Thereby, highly accurate alignment can be performed automatically. Further, by obtaining the correlation between the third attention area and the fourth attention area for each of a plurality of locations, and obtaining the optimum correlation from the obtained correlation, it is possible to perform more precise local alignment. .

  ADVANTAGE OF THE INVENTION According to this invention, the medical image processing apparatus which produces a long image or a difference image efficiently can be provided by performing highly accurate position alignment automatically.

  The best mode for carrying out the present invention will be described below with reference to the accompanying drawings.

<First Embodiment>
The present invention uses an X-ray apparatus, which is one of the medical image capturing apparatuses 2, to join two images captured by moving the imaging position in the body axis direction of the subject so as to partially overlap each other. A long image is created.

  FIG. 1 is a hardware configuration diagram showing the overall configuration of the medical image processing apparatus according to the first embodiment of the present invention.

  The medical image processing apparatus 10 includes an X-ray apparatus, an X-ray CT apparatus, and the like, a medical image capturing apparatus 2 that captures an image of the subject, and an image database in which images of the subject captured by the medical image capturing apparatus 2 are stored. 4 and a network such as LAN3.

  The medical image processing apparatus 10 mainly includes a central processing unit (CPU) 11 as a control device that controls the operation of each component, and a main memory 12 that stores a control program for the device and serves as a work area when the program is executed. And a magnetic disk 13 in which various application software including a program for processing such as an operating system (OS), a device drive of a peripheral device, and measurement of the thickness of the chest wall are stored, and display data is temporarily stored Display memory 14, a monitor 15 such as a CRT monitor or a liquid crystal monitor that displays an image based on data from the display memory 14, a mouse 17 as a position input device, and the state of the mouse 17 is detected and is displayed A controller 1 that outputs signals such as the position of the mouse pointer and the state of the mouse 17 to the CPU 11 When, a keyboard 16 for operator inputs an instruction, and a bus 19 which connects the above components.

  The CPU 11 reads the program from the magnetic disk 13, loads it into the main memory 12, and executes it.

  In this embodiment, the magnetic disk 13 is connected as a storage device other than the main memory 12, but a hard disk drive or the like may be connected in addition thereto.

  Next, a method for creating a long image by joining two images captured by moving the imaging position in the body axis direction of the subject so as to partially overlap each other will be described.

  FIG. 2 is a flowchart showing a processing flow of the medical image processing apparatus 10. The CPU 11 operates according to this flowchart. In the following processing, two or more pieces of image data obtained by moving the imaging position in the body axis direction of the subject so as to partially overlap are read from the medical image imaging apparatus 2 to the medical image processing apparatus 10. Will start after. Note that the processing is started after two or more pieces of image data taken along the body axis direction of the subject stored in the image database 4 are read from the image database 4 into the medical image processing apparatus 10. It may be. Note that the image data read by the medical image processing apparatus 10 is temporarily stored in the magnetic disk 13.

  First, two images 20 and 21 that are stored on the magnetic disk 13 and are imaged by moving the imaging position in the body axis direction of the subject so as to partially overlap each other, and the incidental attached to the images. Information is read into the main memory 12 (step S10). The incidental information includes, for example, the distance traveled by the detector of the X-ray apparatus when the second image is captured after the first image is captured, and the body of the overlapping region of the two images 20 and 21. Information on the length in the axial direction, the date and time when the image was taken, and the like are included.

  Next, a process of extracting an area (calculation target area) used in the alignment process is performed from the two read images 20 and 21 (steps S12 to S18).

  First, a threshold value is determined (step S12). The threshold value is set in advance so that a region used in the alignment process is clearly separated. Note that the threshold value may be automatically set by searching for a lower limit value and an upper limit value with the density indicating the highest frequency in the density range of the X-ray image taken as a median value.

  A binarized image is created by performing threshold processing on the images 20 and 21 based on the threshold determined in step S12. (Step S14).

  A label image is created by performing a two-dimensional labeling process on the binarized image created in step S14 (step S16).

  In the label image created in step S16, the label area having the maximum area is extracted as a calculation target area (step S18).

  Thereby, as shown in FIG. 3, the calculation target region 30 is set to the images 20 and 21. In this embodiment, the spine is processed as the calculation target area.

  Next, rough alignment is performed on the images 20 and 21 (steps S20 to S28). Hereinafter, rough alignment will be described.

  The center-of-gravity coordinates of the calculation target region 30 extracted in step S18 and the width of the extraction region are obtained (step S20). Based on the result, the center-of-gravity coordinates of the calculation target region 30 are obtained as shown in FIG. A template region 31 is set in the image (lower image) 20 centering on (step S22). Note that the size of the template region 31 is set in advance according to the size of the calculation target region 30.

  As for the image (upper image) 21, as shown in FIG. 4B, the search area 32 is set in the image 21 around the center of gravity coordinates of the calculation target area 30 (step S <b> 24). The search area 32 is set to be slightly larger than the size of the template area 31 (about 1 to 2 times the template area 31 and preset in accordance with the size of the template area 31).

  Thereafter, between the template region 31 set in step S22 and the search region 32 set in step S24, the calculation result at the place with the highest correlation is calculated as the movement distance by the following equation (1) ( Step S26).

    Note that the equation for obtaining the correlation is not limited to the above equation (1), and various other evaluation functions such as the function shown in equation (2) can be used.

 Then, the images 20 and 21 are moved using the movement distance obtained in step S26 (step S28).

  As a result, the rough alignment is completed by the first attention area, the template area as the second attention area, and the search area, and the images 20 and 21 are roughly aligned as shown in FIG. An image 22 is created. At this stage, the alignment in the lateral direction (the x direction perpendicular to the y direction which is the body axis direction) is almost completed.

  Next, alignment processing (local alignment) using the shape feature amount is performed on the image 22 in which the images 20 and 21 are roughly aligned (steps S30 to S48). Here, the shape feature means continuity of the shape of the contour shown in the image. In the present embodiment, the alignment means focusing on the continuity of the shape of the spine (spine). Hereinafter, an alignment process (local alignment) using the shape feature amount will be described.

  First, as shown in FIG. 5A, a processing range 33 for performing the alignment process is set (step S30). The processing range 33 includes an overlapping region 22a that is a portion where the images 20 and 21 are pasted, and has a predetermined width in the y direction. Note that the width in the y direction of the processing range 33 is a size necessary for the processing described below to be accurately performed, and is set in advance.

  At a position Y in the y direction in the processing range 33, the densities of all the pixels Y 'in the x direction are added (step S32). As a result, as shown in FIG. 5B, a density value 41 in Y is calculated.

  Thereafter, it is determined whether or not the density addition processing performed in step S32 has been performed on the entire processing range 33 (step S34).

  If the density addition process has not been performed on the entire processing range 33 (NO in step S34), the process returns to step S32, and the density addition process is performed again.

  When density addition processing has been performed on all the processing ranges 33 (YES in step S34), a broken line 40 of density values is obtained as shown in FIG. The length of the part (spine) is measured (step S36). Here, the length is the distance from the contour 42 (the portion where the density value of the broken line 40 is high) to the contour 42, that is, the length of the spine. When the distance between the contour 42 and the contour 42 is short, the length is not measured.

  Since the density of the end face of the spine is higher than that of the other part, the added value when the end face of the spine is included is higher than the added value of the other part. Therefore, the widths 43α, 43β, 43γ, and 43δ in the y direction between the contour 42 and the contour 42 mean the lengths of the spines 34α, 34β, 34γ, and 34δ (see FIG. 5A), respectively.

  In this embodiment, for the sake of explanation, the case where the length is measured for four vertebrae is shown, but the number of vertebrae to be measured is not limited to this.

  Next, the y value and length of the site of interest (spine) are obtained (step S38). Here, the y value is a value of the y axis when the upper left (intersection of the x axis and the y axis) of the image 22 (see FIG. 5A) is 0. For example, as shown in FIGS. 5A and 5B, the y value of the spine 34α is the median value yα of the spine 34α in the y direction, and the length is 43α. Similarly, the relationship as shown in Table 1 is obtained by performing the above-described processing on all the attention sites (vertebrae) included in the processing range 33.

   When the processing is completed for all the target regions (vertebrae), the measured length change is obtained by correlating the y value and the length of the target region (spine) (step S40). That is, as shown in FIG. 6, the values in Table 1 are plotted on a graph in which the vertical direction (body axis direction) is the y value and the horizontal axis is the length, and these are connected by interpolation approximate lines by the least square method or the like. Thus, an approximate line 50 (solid line in FIG. 6) indicating the measured length change is created.

  Thereafter, an ideal length change is obtained from the y value and length of the site of interest (spine) (step S42). The length, that is, the length of the spine is characterized by being equal or longer in the direction from the head to the foot. That is, as shown in FIG. 6, the values in Table 1 are plotted on a graph with the y-axis on the vertical axis and the length on the horizontal axis, and the approximate curve includes as many plots as possible. By creating an approximate straight line that increases or becomes equivalent in length, an approximate line 51 (dotted line in FIG. 6) indicating an ideal length change is created.

  Thereby, an approximate line indicating a change in length, that is, a value related to the continuity of the shape of the spine is obtained.

  When the approximate line 50 indicating the measured length change and the approximate line 51 indicating the ideal length change are created, does the approximate line 50 indicating the measured length change match the approximate line 51 indicating the ideal length change? It is determined whether or not (step S44).

  When the approximate line 50 indicating the measured length change does not match the approximate line 51 indicating the ideal length change, the approximate line 50 indicating the measured length change and the approximate line 51 indicating the ideal length change. The difference (shift amount) is calculated (step S46), and the image 20 or the image 21 is moved so that the shift amount disappears (step S48). For example, in FIG. 6, the length of the approximate line 50 indicating the measured length change is shifted in the direction of shortening around the spine 43γ, compared to the approximate line 51 indicating the ideal length change. The image 20 or the image 21 is moved so as to become longer by the shift amount.

  When the approximate line 50 indicating the measured length change matches the approximate line 51 indicating the ideal length change, it is determined that the image 20 and the image 21 are accurately aligned. The registration process using the quantity is terminated.

  Thereby, the local alignment using the shape feature amount is completed, and the images 20 and 21 are not only in the horizontal direction (x direction perpendicular to the y direction which is the body axis direction) but also in the vertical direction (y direction). Complete alignment is completed.

  Next, density value correction processing (steps S50 to S54) is performed on the portion where the images 20 and 21 are bonded. Hereinafter, the density value correction process will be described.

  Even when shooting is performed under the same conditions, the densities of the image 20 and the image 21 are generally different. For this reason, the overlapping regions 22a (see FIG. 5A) of the images 20 and 21 and the regions 22b and 22c that are not the overlapping regions 22a are discontinuous in density, resulting in an unnatural image. Therefore, in order to join the image 20 and the image 21 more smoothly, it is necessary to perform density composition processing of the overlapping region 22a based on the density values of the images 20 and 21.

  As shown in FIG. 7, the density value of the pixel of interest is calculated by Equation (3) according to the distance between the pixels of interest in the y direction from the pasted portion (step S50).

  When the density value of the pixel of interest is calculated, the density value of the image 22 is corrected in the y direction from the upper left pixel of the image 22 (step 52), and this density value correction is performed on the entire image 22 It is determined whether or not (step S54). When it is determined that the density correction process has been performed on the entire image 22, the process ends.

  As a result, the density of the image 22 that has been aligned is corrected, the images 20 and 21 are smoothly joined, and a natural long image 23 (see FIG. 8A) is generated.

  When the long image 23 is created, it is displayed on the monitor 15 via the display memory 14 (step S56). In addition to the image 23, the monitor 15 has a cross key 56 (FIG. 8) for enlarging and reducing the image 23 and moving the image 23 by an operator operating the keyboard 16 and the mouse 17. (B)), a mode screen 58 for selecting a mode such as color display, black and white display (see FIG. 8C), a cross key 56, a cursor 56 for operating the mode screen 58, and the like are displayed. Can do.

  According to the present embodiment, a template region that is a first region of interest set for two images that are imaged by moving the imaging position in the body axis direction of the subject so as to partially overlap each other The first image and the second image are roughly aligned based on the correlation between the image and the search area, which is the second region of interest, and automatic alignment is performed based on the result. Therefore, highly accurate alignment can be performed.

  Also, use the value obtained from the continuity of the shape of the spine (spine), which is the shape feature quantity, that is, the value obtained from the feature that the length of the spine is equal or longer as it goes from the head to the foot. Thus, alignment in the body axis direction can be performed precisely.

  Further, density correction is performed on the image that has been aligned, so that the two images are smoothly joined, and as a result, a natural long image (third image) can be generated. . Since a long image is automatically generated, a long image can be created easily and efficiently.

  In this embodiment, the method of joining two images has been described. However, even when there are three or more images, the images can be joined by performing the same processing.

<Second Embodiment>
In the medical image processing apparatus according to the first embodiment, local alignment is performed based on the shape feature amount with respect to the two images roughly aligned, but the local alignment is performed. The method to be performed is not limited to this.

  The medical image processing apparatus according to the present embodiment performs local alignment on two roughly aligned images by correlation of a plurality of regions of interest. FIG. 9 is a flowchart showing a processing flow of the medical image processing apparatus 10a according to the second embodiment of the present invention. In the figure, the same portions as those of the first embodiment are denoted by the same reference numerals, and description thereof is omitted.

  First, the two images 20 and 21 stored on the magnetic disk 13 and the accompanying information attached to the images are read into the main memory 12 (step S10).

  Next, a process of extracting an area (calculation target area) used in the alignment process is performed from the two read images 20 and 21 (steps S12 to S18).

  Next, rough alignment is performed on the images 20 and 21 (steps S20 to S28).

  In addition, since the process of step S10-S28 is the same as 1st Embodiment, detailed description is abbreviate | omitted.

  When the rough alignment shown in steps S20 to S28 is completed, the alignment processing 1 (local alignment) using the density feature amount is performed on the roughly aligned images 20 and 21 (step alignment). S60 to S76). Here, the density feature amount refers to a feature obtained by performing threshold processing on the density of an image. Hereinafter, the registration process 1 (local registration) using the density feature amount will be described.

  As shown in FIGS. 10A-1 and 10A-2, a template area 35-n is set inside the search area 32 set in the image (upper image) 21 in step S24 (step S60). ).

  As for the image (lower image) 20, as shown in FIGS. 10 (b-1) and (b-2), a search is performed inside the template region 31 set in the image (lower image) 20 in step S22. Region 36 is set (step S62).

  In the present embodiment, the height of the template region 35-n is about 20 mm, and the height of the search region 36 is about 40 mm. The template region 35-n and the search region 36 are set so as to include the space between the vertebrae.

  First, template area 35-n is set such that n = 1, that is, template area 35-1 is positioned at the left end of search area 36, as shown in FIGS. 10 (a-2) and 10 (b-2). (Step S64).

  Then, the movement distance in the template area 35-n is calculated by moving the template area 35-n up and down within the search area 36 (step S66) (step S68). In the present embodiment, the height of the template region 35-n is about 20 mm, and the height of the search region 36 is about 40 mm. Therefore, the template region 35-n is about 20 mm up and down inside the search region 36. Can move. The movement distance can be calculated using the same method as in step S26.

  Then, it is determined whether n = N (step S72).

  If n = N is not satisfied (NO in step S72), n = n + 1 is set (step S70), and then the process returns to step S66, where the template area 35-n is set again. For example, when n = 1 in step S68, the template area 35-1 is located at the left end of the search area 36 as shown in FIGS. 10 (a-2) and (b-2). When n = 2 is set in S70, as shown in FIGS. 10A-3 and 10B-3, the template area 35- is moved to the position moved to the right by one pixel from the position of the template area 35-1. 2 is set.

  When n = N (YES in step S72), that is, when the template area 35-N is located at the right end of the search area 36 as shown in FIGS. 10 (a-4) and (b-4). The most frequent travel distance (the most frequent travel distance) among the plurality of travel distances obtained in step S72 is calculated (step S74). The calculation of the most frequent movement distance may be performed by creating a histogram of the plurality of movement distances obtained in step S68 and calculating the median value of the class having the highest frequency as the most frequent movement distance, or in step S68. It is also possible to create a histogram of a plurality of movement distances obtained in step 1, calculate the center of gravity of the histogram, and calculate the value as the most frequent movement distance.

  Then, the images 20 and 21 are moved using the most frequent moving distance obtained in step S74 (step S76).

  Thereby, the registration using the density feature amount by the third attention area, the template area which is the fourth attention area, and the search area is finished, and the images 20 and 21 are locally aligned, and the horizontal direction (body The precise alignment not only in the axial direction (x direction perpendicular to the y direction) but also in the vertical direction (y direction) is completed.

  When the registration process 1 (local registration) using the density feature amount is completed, the density value correction process (steps S50 to S54) of the part where the images 20 and 21 are bonded together, and the display of the generated long image (Step S56). Is done. In addition, since the process of step S50-S56 is the same as 1st Embodiment, detailed description is abbreviate | omitted.

  According to the present embodiment, the correlation between a plurality of template regions that are third attention regions and a plurality of fourth attention regions that are extracted from a portion where rough positioning has been performed. Based on the above, by performing the local alignment of the two images, it is possible to automatically perform the high-accuracy alignment.

  Further, by obtaining a correlation with a search area for a certain template area and obtaining an optimum correlation from a plurality of obtained correlations (movement distances), the deviation between the two images is 1 mm or less. Precise local alignment can be performed.

<Third Embodiment>
In the medical image processing apparatus according to the second embodiment, the template region is moved pixel by pixel in the search region for two roughly aligned images, and the movement distance is measured at each location. In addition, the local alignment is performed by calculating the most frequently moved distance among them, but the method of performing the local alignment is not limited to this.

  The medical image processing apparatus according to the present embodiment sets a plurality of template regions and search regions for two roughly aligned images, measures the moving distance in each of them, and most frequently among them. Local alignment is performed by calculating the movement distance. FIG. 11 is a flowchart showing a processing flow of the medical image processing apparatus 10b according to the second embodiment of the present invention. In the figure, the same portions as those in the first embodiment and the second embodiment are denoted by the same reference numerals, and description thereof is omitted.

  First, the two images 20 and 21 stored on the magnetic disk 13 and the accompanying information attached to the images are read into the main memory 12 (step S10).

  Next, a process of extracting an area (calculation target area) used in the alignment process is performed from the two read images 20 and 21 (steps S12 to S18).

  Next, rough alignment is performed on the images 20 and 21 (steps S20 to S28).

  In addition, since the process of step S10-S28 is the same as 1st Embodiment, detailed description is abbreviate | omitted.

  When the rough alignment shown in Steps S20 to S28 is completed, alignment processing 2 (local alignment) using density feature amounts is performed on the roughly aligned images 20 and 21 (Step S20). S80 to S90). Hereinafter, the alignment process 2 (local alignment) using the density feature amount will be described.

  As shown in FIGS. 12A-1 and 12A-2, a plurality of template regions 37 are set inside the template region 31 set in the image (lower image) 20 in step S22 (step S80). ).

  As for the image (upper image) 21, as shown in FIGS. 12B-1 and 12B-2, the template inside the search area 32 set in the image (upper image) 21 in step S24. A search area 38 is set at a location corresponding to the area 37 (step S82).

  For example, as shown in FIGS. 12A-2 and 12B-2, for the template area 37-1, a search area 38-1 is set at the corresponding location. Similarly, search areas 38-1 to 38-9 are set for template areas 37-1 to 37-9. The template areas 37-1 to 37-9 and the search areas 38-1 to 38-9 are set to be smaller than the template area 31 and the search area 32 based on the movement distance obtained in step S26. . In FIG. 12, nine template areas 37-1 to 37-9 and search areas 38-1 to 38-9 are set. However, if there are a plurality of template areas 37 and search areas 38, the number is limited to nine. Absent.

  When the setting of the template area 37 and the search area 38 is completed, the movement distance in each template area 37 and the search area 38 is calculated (step S84). That is, the movement distance is calculated between the template area 37-1 and the search area 38-1, and the movement distance between the template area 37-1 to 37-9 and the search area 38-1 to 38-9 is calculated in the same manner. Calculated. The movement distance can be calculated using the same method as in step S26.

  Thereafter, it is determined whether the movement distance has been calculated for all template regions 37 and search regions 38 (step S86). In the present embodiment, nine template areas 37-1 to 37-9 and nine search areas 38-1 to 38-9 are set for the template area 37 and the search area 38, respectively. It is determined whether the distance has been calculated.

  If the movement distance is not calculated for all the template areas 37 and the search areas 38, the process returns to step S84, and the movement distance is calculated again.

  When the movement distances are calculated for all the template areas 37 and the search areas 38, the most frequent movement distance (the most frequent movement distance) among the plurality of movement distances obtained in step S86 is calculated. (Step S88). The calculation of the most frequent movement distance may be performed by creating a histogram of the plurality of movement distances obtained in step S86 and calculating the median value of the class having the highest frequency as the most frequent movement distance. It is also possible to create a histogram of a plurality of movement distances obtained in step 1, calculate the center of gravity of the histogram, and calculate the value as the most frequent movement distance.

  Then, the images 20 and 21 are moved using the most frequently moved distance obtained in step S88 (step S90).

  Thereby, the registration using the density feature amount by the third attention area, the template area which is the fourth attention area, and the search area is finished, and the images 20 and 21 are locally aligned, and the horizontal direction (body The precise alignment not only in the axial direction (x direction perpendicular to the y direction) but also in the vertical direction (y direction) is completed.

  When the registration process 2 (local registration) using the density feature amount is completed, the density value correction process (steps S50 to S54) of the part where the images 20 and 21 are bonded together, and the display of the generated long image (Step S56). Is done. In addition, since the process of step S50-S56 is the same as 1st Embodiment, detailed description is abbreviate | omitted.

  According to the present embodiment, the correlation between a plurality of template regions that are third attention regions and a plurality of fourth attention regions that are extracted from a portion where rough positioning has been performed. Based on the above, by performing the local alignment of the two images, it is possible to automatically perform the high-accuracy alignment.

<Fourth embodiment>
In the medical image processing apparatus according to the first embodiment, the local alignment using the shape feature amount is performed after the rough alignment using the density feature amount. However, the present invention is not limited to this. is not.

  The medical image processing apparatus according to the present embodiment performs local alignment by two methods and compares the results. FIG. 13 is a flowchart showing a processing flow of the medical image processing apparatus 10c according to the fourth embodiment of the present invention. In the figure, the same portions as those in the first embodiment, the second embodiment, or the third embodiment are denoted by the same reference numerals, and the description thereof is omitted.

  First, the two images 20 and 21 stored on the magnetic disk 13 and the accompanying information attached to the images are read into the main memory 12 (step S10).

  Next, a process of extracting an area (calculation target area) used in the alignment process is performed from the two read images 20 and 21 (steps S12 to S18).

  Next, rough alignment is performed on the images 20 and 21 (steps S20 to S28).

  In addition, since the process of step S10-S28 is the same as 1st Embodiment, detailed description is abbreviate | omitted.

  When the images 20 and 21 are roughly aligned, local alignment is performed on the images 20 and 22 on which the images 20 and 21 are roughly aligned (steps S30 to S96). Hereinafter, local alignment will be described.

  First, alignment processing (local alignment 1) using shape feature values is performed (steps S30 to S47).

  First, a processing range 33 for performing alignment processing is set (step S30).

  At a position Y in the y direction in the processing range 33, the densities of all the pixels Y ′ in the x direction are added (step S32), and whether or not this density addition processing has been performed on all the processing ranges 33. Determination is made (step S34).

  If the density addition process has not been performed on the entire processing range 33 (NO in step S34), the process returns to step S32, and the density addition process is performed again.

  When the density addition processing has been performed on the entire processing range 33 (YES in step S34), the length of the site of interest (spine) is measured (step S36).

  Next, the y value and length of the site of interest (spine) are obtained (step S38).

  When the processing is completed for all the target regions (vertebrae), the measured length change is obtained by correlating the y value and the length of the target region (spine) (step S40).

  Thereafter, an ideal length change is obtained from the y value and length of the site of interest (spine) (step S42).

  When the approximate line 50 indicating the measured length change and the approximate line 51 indicating the ideal length change are created, does the approximate line 50 indicating the measured length change match the approximate line 51 indicating the ideal length change? It is determined whether or not (step S44).

  In addition, since the process of step S30-S44 is the same as 1st Embodiment, detailed description is abbreviate | omitted.

  When the approximate line 50 indicating the measured length change does not match the approximate line 51 indicating the ideal length change, the approximate line 50 indicating the measured length change and the approximate line 51 indicating the ideal length change. The difference (deviation amount) is calculated, and this deviation amount (see FIG. 6) is set as the movement amount p in the alignment using the shape feature quantity (step S47).

  When the approximate line 50 indicating the measured length change matches the approximate line 51 indicating the ideal length change, it is determined that the image 20 and the image 21 are accurately aligned. The movement amount p in the alignment using the amount is set to zero.

  Thereby, the alignment process (local alignment 1) using the shape feature amount shown in steps S30 to S47 is completed.

  Next, alignment processing 1 (local alignment 2) using density feature amounts is performed on the roughly aligned images 20 and 21 (steps S60 to S75).

  A template area 35-n is set inside the search area 32 set in the image (upper image) 21 in step S24 (step S60). Also for the image (lower image) 20, the search area 36 is set inside the template area 31 set in the image (lower image) 20 in step S22 (step S62).

  After the template area 35-n and the search area 36 are set, n = 1 is set (step S64).

  Then, the movement distance in the template area 35-n is calculated by moving the template area 35-n up and down within the search area 36 (step S66) (step S68).

  Then, it is determined whether n = N (step S72).

  If n = N is not satisfied (NO in step S72), n = n + 1 is set (step S70), and then the process returns to step S66, where the template area 35-n is set again.

  In addition, since the process of step S60-S70 is the same as 2nd Embodiment, detailed description is abbreviate | omitted.

  In the case of n = N (YES in step S72), the most frequent travel distance (most frequent travel distance) is calculated among the plurality of travel distances obtained in step S68, and this most frequent travel distance is the shape. It is set as the movement amount q in the alignment using the feature amount (step S75).

  The calculation of the most frequent movement distance may be performed by creating a histogram of the plurality of movement distances obtained in step S68 and calculating the median of the class with the highest frequency as the most frequent movement distance. A histogram of a plurality of movement distances obtained in step S68 may be created, the center of gravity of the histogram may be calculated, and the value may be calculated as the most frequent movement distance.

  Thereby, the registration process 1 (local registration 2) using the density feature amount shown in steps S60 to S75 is completed.

  When the registration processing (local registration 1) using the shape feature amount and the registration processing 1 (local registration 2) using the density feature amount are completed, the shape feature amount obtained in step S47 is obtained. The amount of movement p in the used alignment processing (local alignment 1), and the amount of movement q in the alignment processing 1 (local alignment 2) using the density feature obtained in step S75. It is determined whether or not the difference is equal to or less than a threshold (for example, one pixel) (step S92). The threshold value is set in advance, but can be changed by an operator.

  When the difference between the movement amount p and the movement amount q is equal to or smaller than the threshold value (in the case of YES in step S92), the movement amount q is set as the movement amount in the local alignment (step S94), and the image 20, 21 is moved (step S96).

  If the difference between the movement amount p and the movement amount q is not less than or equal to the threshold value (NO in step S92), the process returns to step S20, and the rough alignment is performed again. Alignment 1) and alignment processing 1 (local alignment 2) using density feature amounts are performed. Note that the template area, search area, or processing range set in this case is set to be smaller than the template area, search area, or processing range that is used for the first time.

  As a result, the local alignment shown in steps S30 to S96 is completed, and the images 20 and 21 are accurate not only in the horizontal direction (x direction perpendicular to the y direction which is the body axis direction) but also in the vertical direction (y direction). Alignment is performed.

  In the present embodiment, when the difference between the movement amount p and the movement amount q is less than or equal to the threshold value, the images 20 and 21 are moved using the movement amount q. Alternatively, the movement amount p and the movement amount q may be compared, and the smaller one may be used.

  When the local alignment is completed, density value correction processing (steps S50 to S54) is performed on the portion where the images 20 and 21 are bonded, and the image is displayed on the monitor (step S56). In addition, since the process of step S50-S56 is the same as 1st Embodiment, detailed description is abbreviate | omitted.

  According to the present embodiment, there are two types of alignment processing (local alignment 1) using shape feature amounts and alignment processing 1 (local alignment 2) using density feature amounts. By performing local alignment by the method and confirming whether or not the difference is equal to or less than a predetermined threshold value, the local alignment can be performed more accurately and more reliably.

<Fifth embodiment>
In the medical image processing apparatus according to the first embodiment, first, rough alignment is performed on two images captured by moving the imaging position in the body axis direction of the subject so that a part of the medical image processing apparatus overlaps. However, the present invention is not limited to this.

  The medical image processing apparatus according to the present embodiment accurately aligns a tomographic image continuously taken in the body axis direction of a subject by first roughly aligning and then performing local alignment. Alignment is performed. FIG. 14 is a flowchart showing a process flow of the medical image processing apparatus 10d according to the fifth embodiment of the present invention. In the figure, the same portions as those of the first embodiment are denoted by the same reference numerals, and description thereof is omitted.

  In the following processing, a tomographic image of a subject imaged by an X-ray CT apparatus, which is one of the medical image capturing devices 2, and a tomographic image of the same subject imaged in the past stored in the image database 4 are medically used. The process is started after reading into the image processing apparatus 10. Note that it may be started after the tomographic images of the same subject having different imaging timings are read from the image database 4 to the medical image processing apparatus 10. Note that the image data read by the medical image processing apparatus 10 is temporarily stored in the magnetic disk 13.

  First, alignment processing (rough alignment 2) in the body axis direction is performed on two sets of tomographic images En and Fn (n = 1 to N, see FIG. 15) photographed at different times ( Steps S100 to S152). In this embodiment, the bronchial bifurcation is identified from two sets of tomographic images taken at different times, and the positions of the bronchial bifurcation of the tomographic image taken in the past and the bronchial bifurcation of the recently taken tomographic image are determined. This is a process of matching the positions in the body axis direction of two sets of tomographic images taken at different times by matching. Hereinafter, the rough alignment 2 will be described.

  First, processing for specifying a slice to be aligned in the body axis direction is performed on a tomographic image (past image) En (n = 1 to N, see FIG. 15A) of a subject imaged in the past. (Steps S100 to S124).

  First, feature area extraction processing is performed (steps S100 to S114). Hereinafter, the feature region extraction process will be described.

  The past image En (n = 1 to N) and the incidental information attached to the image stored in the magnetic disk 13 are read into the main memory 12 (step S100). The incidental information includes, for example, information on the date and time when the past image En was captured.

  n = 1 is set (step S102). That is, the process starts from the first slice E1 of the past image En.

  Thereafter, a threshold value is determined (step S104). This threshold value is set in advance to an optimum value that clearly separates the area used in the alignment process. Note that the threshold value may be automatically set by searching for the lower limit value and the upper limit value with the CT value indicating the highest frequency within the range of CT values included in the past image En as the median value.

  Then, based on the threshold determined in step S104, threshold processing is performed on the past image En, thereby creating a binary image of the past image En (step S106).

  A two-dimensional labeling process is performed on the binarized image created in step S106, and a body region is created by using the label region of the maximum area (step S108).

  An air region in the body region is extracted by performing a logical operation on the body region created in step S108 and the binarized region (step S110).

  An isolated region that is near the center of the body region with respect to the air region created in step S110 is extracted as a feature region (step S112). Here, the characteristic regions are the trachea and the bronchial region.

  Then, it is determined whether n = N, that is, whether the above processing has been completed for all slices of the past image En (step S114).

  If feature regions have not been extracted for all slices (NO in step S114), n = N + 1 is set (step S115), and the process returns to step S104. That is, threshold setting (step S104) is performed for the next slice En.

  If feature regions have been extracted for all slices (YES in step S114), the feature region extraction process ends.

  When the feature region extraction process is completed, a slice having a feature in the feature region, that is, a slice having a bronchial bifurcation where the bronchial region that is the feature region branches is identified (steps S116 to S124).

  n = 1 is set (step S116). That is, the process starts from the first slice E1 of the past image En.

  It is determined whether there is one feature region (step S118). That is, it is determined whether one bronchus is included in the slice.

  If there is not one feature region (NO in step S118), n = n + 1 is set (step S119), and the process returns to step S118. That is, the process proceeds to the next slice, and it is determined again whether there is one feature region (step S118).

  If there is one feature region (YES in step S118), it is determined whether the feature region of slice En and the feature region of slice En + 1 are continuous (step S120). That is, it is determined whether the bronchus included in the slice En and the bronchus included in the slice En + 1 are continuous.

  If the feature region of slice En and the feature region of slice En + 1 are not continuous (NO in step S120), n = n + 1 is set (step S119), and the process returns to step S118. That is, the process proceeds to the next slice, and it is determined again whether there is one feature region (step S118).

  If the feature region of slice En and the feature region of slice En + 1 are continuous (YES in step S120), it is determined whether there are two feature regions of slice En + 1 (step S122). That is, it is determined whether or not there is two slices in slice En + 1, that is, one bronchus branch portion where the bronchus is branched.

  If there are not two feature regions in the slice En + 1 (NO in step S122), n = n + 1 is set (step S119), and the process returns to step S118. That is, the process proceeds to the next slice, and it is determined again whether there is one feature region (step S118).

  If there are two feature regions of the slice En + 1 (YES in step S122), the slice En is identified as a slice to be aligned in the past image (step S124).

  Thereby, in the past image En, a slice to be aligned, that is, a slice including a bronchial bifurcation is specified.

  When the process of specifying the slice for positioning in the body axis direction with respect to the past image En is completed, the tomographic image (current image) Fn (n = 1 to N, FIG. 15B) of the subject imaged recently. (Refer to step S126 to S150) for identifying a slice to be aligned in the body axis direction.

  First, feature area extraction processing is performed (steps S126 to S140). Hereinafter, the feature region extraction process will be described.

  The current image Fn (n = 1 to N) and the incidental information attached to the image stored in the magnetic disk 13 are read into the main memory 12 (step S126). The incidental information includes, for example, information on the date and time when the current image Fn was taken.

  n = 1 is set (step S128). That is, the processing starts from the first slice F1 of the current image Fn.

  Thereafter, a threshold value is determined (step S130). This threshold value is set in advance to an optimum value that clearly separates the area used in the alignment process. The threshold value may be automatically set by searching for the lower limit value and the upper limit value with the CT value indicating the most frequent frequency within the range of CT values included in the current image Fn as the median value.

  Then, based on the threshold value determined in step S130, threshold processing is performed on the current image Fn to create a binary image of the current image Fn (step S132).

  A two-dimensional labeling process is performed on the binarized image created in step S132, and a body region is created by using the label region of the maximum area (step S134).

  An air region in the body region is extracted by performing a logical operation on the body region created in step S134 and the binarized region (step S136).

  An isolated region that is near the center of the body region with respect to the air region created in step S136 is extracted as a feature region (step S138). Here, the characteristic regions are the trachea and the bronchial region.

  Then, it is determined whether n = N, that is, whether the above processing has been completed for all slices of the current image Fn (step S140).

  If feature regions have not been extracted for all slices (NO in step S140), n = N + 1 is set (step S141), and the process returns to step S130. That is, threshold setting (step S130) is performed for the next slice Fn.

  If feature regions have been extracted for all slices (YES in step S140), the feature region extraction process ends.

  When the feature region extraction processing is completed, a slice having a feature in the feature region, that is, a slice having a bronchial bifurcation where the bronchial region that is the feature region branches is identified (steps S142 to S150).

  n = 1 is set (step S142). That is, the processing starts from the first slice F1 of the current image Fn.

  It is determined whether there is one feature region (step S144). That is, it is determined whether one bronchus is included in the slice.

  If there is not one feature region (NO in step S144), n = n + 1 is set (step S145), and the process returns to step S144. That is, the process proceeds to the next slice, and it is determined again whether there is one feature region (step S144).

  When there is one feature region (YES in step S144), it is determined whether the feature region of slice Fn and the feature region of slice Fn + 1 are continuous (step S146). That is, it is determined whether the bronchus included in the slice Fn and the bronchus included in the slice Fn + 1 are continuous.

  When the feature region of slice Fn and the feature region of slice Fn + 1 are not continuous (NO in step S146), n = n + 1 is set (step S145), and the process returns to step S144. That is, the process proceeds to the next slice, and it is determined again whether there is one feature region (step S144).

  If the feature region of slice Fn and the feature region of slice Fn + 1 are continuous (YES in step S146), it is determined whether there are two feature regions of slice Fn + 1 (step S148). That is, it is determined whether or not there is two slices in the slice Fn + 1 in the slice Fn, that is, whether or not it is a bronchial bifurcation where the bronchus is branched.

  If there are not two feature regions in the slice Fn + 1 (NO in step S148), n = n + 1 is set (step S145), and the process returns to step S144. That is, the process proceeds to the next slice, and it is determined again whether there is one feature region (step S144).

  If there are two feature regions of slice Fn + 1 (YES in step S148), this slice Fn is identified as a slice to be aligned in the current image (step S150).

  Thereby, in the current image Fn, a slice to be aligned, that is, a slice including a bronchial bifurcation is specified.

  When the past image En and the current image Fn to be aligned are specified, the past image En and the current image Fn are aligned in the body axis direction (step S152). For example, as shown in FIG. 15, in the current images F1 to FN, the slice corresponding to the past image E3 (chest) is F3, and the slice corresponding to the past image E20 (buttock) is F20, so that the past image En And the position of the slice in the body axis direction of the current image Fn.

  Thereby, the alignment process (rough alignment) of the past image En and the current image Fn in the body axis direction is completed.

  When the alignment process (rough alignment 2) between the past image En and the current image Fn in the body axis direction is completed, the local alignment (the alignment of the past image En and the current image Fn) is performed using the density feature amount. Steps S154 to S180) are performed. In the present embodiment, the processing is to align the positions of structures such as organs and blood vessels in the state in which the positions in the body axis direction of two sets of tomographic images taken at different times are aligned. Hereinafter, local registration using the density feature quantity (positioning 3 using the density feature quantity) will be described.

  n = 1 is set (step S154). That is, processing is started from the first slices E1 and F1 of the past image En and the current image Fn.

  First, the image size (the size of the body region) is adjusted using the size and inclination of the body region extracted in step S108 and step S134 (rough alignment 2), and the past image En and the current image Fn are adjusted. The body regions are matched in size (step S156). In the past and present, the size of the body region may not match due to a change in the body shape of the subject. Therefore, by enlarging / reducing the past image En or the current image Fn, the size of the body region is adjusted for each slice, and the size of the body region of the past image En and the size of the body region of the current image Fn are roughly set. Match. Note that various known techniques can be used for enlarging / reducing an image.

  Thereafter, the body regions of the past image En and the current image Fn are divided into a plurality of blocks Bi (i = 2 to I) (step S158). For example, as shown in FIG. 16, with respect to the past image En and the current image Fn, using the air region extracted in step S110 and step S136 (rough alignment 2), the air region and the other region A block Bi including the boundary is set. At this time, the entire body region is covered by the blocks B1 to BI.

  i = 1 is set (step S160).

  A plurality of template regions 60 are set in the block Bi set in step S158 for the past image En (step S162). As shown in FIG. 16A, the template regions 60 are set so as not to overlap each other.

  Similarly, a plurality of search areas 61 are set in the block Bi set in step S158 for the current image Fn (step S164). As shown in FIG. 16B, the search area 61 is set at a position corresponding to the template area 60 so as not to overlap each other.

  For example, as shown in FIG. 16, for the template area 60-1, a search area 61-1 is set at the corresponding location. Similarly, search areas 61-1 to 61-9 are set so as not to overlap each other with respect to template areas 60-1 to 60-9. In FIG. 16, nine template regions 60-1 to 60- and search regions 61-1 to 61-9 are set. However, the number of template regions 60 and search regions 61 is not limited to nine as long as there are a plurality of template regions 60 and search regions 61. .

  In the present embodiment, the template area 60 is set in the past image En and the search area 61 is set in the current image Fn. However, the present invention is not limited thereto, and the search area 61 is set in the past image En and the current image Fn. The template area 60 may be set in

  When the setting of the template area 60 and the search area 61 is completed, the movement distance in each of the template area 60 and the search area 61 is calculated (step S166). That is, the movement distance is calculated between the template area 60-1 and the search area 61-1, and the movement distance between the template areas 60-1 to 60-9 and the search areas 61-1 to 61-9 is calculated in the same manner. Calculated. The movement distance can be calculated using the same method as in step S26 in the first embodiment.

  Thereafter, it is determined whether the movement distance has been calculated for all the template areas 60 and the search areas 61 (step S168). In this embodiment, nine template areas 60-1 to 60-9 and nine search areas 61-1 to 61-9 are set for the template area 60 and the search area 61, respectively. It is determined whether the distance has been calculated.

  If the movement distance is not calculated for all the template areas 60 and the search areas 61, the process returns to step S166, and the movement distance is calculated again.

  When the movement distances are calculated for all the template areas 60 and the search areas 61, the most frequent movement distance (the most frequent movement distance) among the plurality of movement distances obtained in step S166 is calculated. (Step S170). The calculation of the most frequent movement distance may be performed by creating a histogram of the plurality of movement distances obtained in step S170, and calculating the median value of the class having the highest frequency as the most frequent movement distance. It is also possible to create a histogram of a plurality of movement distances obtained in step 1, calculate the center of gravity of the histogram, and calculate the value as the most frequent movement distance.

  Then, it is determined whether i = I, that is, whether the most frequently moved distance is calculated for all the blocks B1 to BI (step S172).

  If the most frequently traveled distance has not been calculated for all the blocks B1 to BI (NO in step S172), i = i + 1 is set (step S174), and then the process returns to step S162. That is, the template area is set for the next block Bi + 1.

  When the most frequent travel distance is calculated for all the blocks B1 to BI (YES in step S172), the most frequent travel distance calculated in a certain block Bi is set as the travel distance in this block Bi. All the blocks B1 to BI are translated (local image deformation processing) by the movement distance in the blocks B1 to BI (step S176).

  Thereby, local alignment using the density feature amount of the past image En and the current image Fn is performed.

  Then, it is determined whether or not n = N, that is, whether or not local alignment (steps S156 to S176) using density feature amounts has been performed for all slices (step S178).

  When local alignment using the density feature amount is not performed for all slices (NO in step S178), n = n + 1 is set (step S180), and then the process returns to step S156. . That is, enlargement / reduction processing is performed on the next slices En + 1 and Fn + 1.

  When local alignment using density feature values is performed for all slices (YES in step S178), local positions using density feature values for all slices. Since the alignment is complete, the local alignment process using the density feature amount is terminated.

  Thereby, the local alignment of the past image En and the current image Fn is completed using the density feature amount.

  When the local alignment between the past image En and the current image Fn is completed, a difference image between the past image En and the current image Fn is created (step S182).

  According to the present embodiment, based on the correlation between the attention areas (bronchial bifurcations) included in the past image and the current image, which are tomographic images continuously taken in the body axis direction of the subject, After roughly aligning the current image in the body axis direction, a plurality of template regions that are third attention regions extracted from the past image subjected to the rough alignment, and extracted from the current image Based on the correlation with the plurality of search areas that are the fourth attention areas, the past image and the current image are automatically aligned in a direction orthogonal to the body axis direction, thereby automatically increasing the height. Accurate alignment can be performed.

  In the present embodiment, the tomographic image taken by the X-ray CT apparatus is aligned and a difference image is created. However, the present invention is not limited to this, and as shown in FIG. Even if a long image parallel to the body axis direction is created from the captured image data by the maximum value projection, the past long image and the current long image are aligned, and a difference image is created. Good.

1 is a schematic diagram showing an overall configuration of a first embodiment of a medical image processing apparatus to which the present invention is applied. It is a flowchart which shows the flow of a process of 1st Embodiment of the said medical imaging device. It is explanatory drawing of the process which extracts the calculation area | region of 1st Embodiment of the said medical imaging device. It is explanatory drawing of the rough alignment of 1st Embodiment of the said medical imaging device. It is explanatory drawing of the positioning (local positioning) using the shape feature-value of 1st Embodiment of the said medical imaging device. It is explanatory drawing of the positioning (local positioning) using the shape feature-value of 1st Embodiment of the said medical imaging device. It is explanatory drawing of density correction of 1st Embodiment of the said medical imaging device. It is an example of the display of 1st Embodiment of the said medical imaging device. It is a flowchart which shows the flow of a process of 2nd Embodiment of the medical image processing apparatus to which this invention was applied. It is explanatory drawing of the positioning 1 (local positioning) using the density | concentration feature-value of 3rd Embodiment of the said medical imaging device. It is a flowchart which shows the flow of a process of 3rd Embodiment of the medical image processing apparatus to which this invention was applied. It is explanatory drawing of the positioning 2 (local positioning) using the density | concentration feature-value of 3rd Embodiment of the said medical imaging device. It is a flowchart which shows the flow of a process of 4th Embodiment of the medical image processing apparatus to which this invention was applied. It is a flowchart which shows the flow of a process of 5th Embodiment of the medical image processing apparatus to which this invention was applied. It is explanatory drawing of the rough alignment 2 of 5th Embodiment of the said medical imaging device. It is explanatory drawing of the position alignment 3 (local position alignment) using the density | concentration feature-value of 5th Embodiment of the said medical imaging device. It is explanatory drawing of the modification of 5th Embodiment of the said medical imaging device.

Explanation of symbols

10: Medical image processing device, 15: Monitor, 20: Lower image, 21: Upper image, 31, 35, 37, 60: Template region, 32, 36, 38, 61: Search region

Claims (3)

Image acquisition means for acquiring a first image in which the subject is imaged and a second image in which a part of the first image overlaps with the first image;
The first region of interest on the first image and the second region of interest on the second image that are used for alignment from the images of overlapping portions when the first image and the second image are aligned First region of interest extraction means for extracting
A correlation between the image in the first region of interest and the image in the second region of interest extracted by the first region of interest extraction unit is obtained, and the first image and the second image are obtained based on this correlation. First alignment means for performing alignment of
Processing area extraction means for extracting a predetermined processing area including an image of the alignment portion from the first image and the second image aligned by the first alignment means;
Second alignment means for aligning the first image and the second image based on the shape feature amount or density feature amount of the image in the extracted processing region;
Image creating means for creating a third image based on the first image and the second image aligned by the second alignment means;
Display means for displaying the third image created by the image creation means;
A medical image processing apparatus comprising:   The medical image processing apparatus according to claim 1, wherein the shape feature amount is a value obtained from continuity of a spine shape. The processing region extraction unit includes a first narrower region than the first region of interest at a plurality of locations in the alignment region between the first image and the second image aligned by the first alignment unit. Extracting a third region of interest on the second image and a fourth region of interest on the second image that is narrower than the second region of interest;
The second alignment means correlates the image in the extracted third region of interest obtained for each of a plurality of locations as the density feature amount and the image in the extracted fourth region of interest. The medical image processing apparatus according to claim 1, wherein alignment of the first image and the second image is performed based on a value related to the image.

Images