JP2006247293A - Image processing method, image processing device, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing method, an image processing device, and a program which fit the position of a subject to be examined between a first image group which consists of two or more images representing the same subject and a second image, and produce a difference image of higher precision irrespective of a change of the posture and condition of the subject when producing the difference picture of the first image group and the second image after positioning. <P>SOLUTION: A positioning means 2 processes a positioning processing between each image P<SB>1</SB>to P<SB>N</SB>of a past image group and an image P<SB>A</SB>for diagnostic reading. A difference calculating means 3 obtains ▵<SB>N</SB>(x, y) from a difference ▵<SB>1</SB>(x, y) with the corresponding pixel in the image P<SB>A</SB>with respect to each pixel of each image P<SB>1</SB>" to P<SB>N</SB>" after positioning processing. A synthetic difference image generation means 4 regards a minimum value of the absolute value of ▵<SB>N</SB>(x, y) from the difference ▵<SB>1</SB>(x, y) with respect to each pixel as a synthetic difference ▵<SB>C</SB>(x, y). A synthetic difference image Q<SB>C</SB>is produced based on this synthetic difference ▵<SB>C</SB>(x, y). This constitution can eliminate the contribution to the synthetic difference image Q<SB>C</SB>of the pixel with the large absolute value of a difference which causes an artifact. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image processing method, apparatus, and program, and more specifically, aligns a subject between a plurality of images representing the same subject, and generates an image representing a difference between the images after the alignment. And a program for causing a computer to perform this method.

  2. Description of the Related Art Conventionally, in various fields, two or more images of the same subject are comparatively read and checked for a difference between both images, and the subject is inspected based on the difference. For example, in the field of manufacturing industrial products, an image taken when a product is new and an image taken after an endurance test of the product are comparatively read and attention is paid to areas where the difference between the two is large. In the medical field, doctors compare and interpret multiple radiographic images taken at different times for a diseased part of a patient in the medical field. By doing so, it is carried out to grasp the progress and cure status of the disease and to examine the treatment policy.

A time-lapse subtraction technique is known as a technique for supporting comparative interpretation of two or more images. The temporal subtraction technique is a technique for performing alignment between a plurality of images representing the same subject at different shooting times, generating a differential image, and extracting and emphasizing a temporally changing portion. Specifically, after performing a rough alignment process (global matching) that detects the overall shift amount (position shift amount) between two images, the shift amount for each local region in both images is detected. Then, local alignment processing (local matching) for determining the shift amount of each pixel is performed based on that, and nonlinear distortion conversion processing (warping) is performed on one image based on the determined shift amount of each pixel. Thus, the subject in both images is aligned. Then, a difference image is generated by performing a difference process between the one image subjected to the nonlinear distortion conversion process and the other image. Thereby, it becomes possible to align two images relatively well (for example, Patent Documents 1, 2, 3, 4, Non-Patent Document 1).
However, this temporal subtraction technique is a technique that realizes alignment between images after being projected in two dimensions, and is a three-dimensional change in the position of the subject at the time of imaging (for example, forward tilt, backward tilt of the patient, It was impossible in principle to deal with (rotation). Also, it was very difficult to cope with a large change in state. For example, when temporal subtraction is applied to a chest radiograph for observing changes in the human chest over time, if there is a large change in lung field size of 1 to 2 cm due to changes in the respiratory phase, A large misalignment artifact (false image) occurs near the diaphragm.

In order to solve such a problem, the applicant of the present application agrees that both images are the same for all combinations of a plurality of past images and a plurality of current images acquired by controlling the patient's posture and respiratory phase. The index value representing the sex is calculated, the combination with the highest matching is extracted, the registration processing is performed on the current image and the past image constituting the combination with the highest matching, and the difference between the two images after the registration is expressed. It is proposed to generate an image (for example, Patent Document 5).
JP-A-7-37074 JP-A-8-335271 JP 2001-157675 A JP 2002-032735 A Japanese Patent Laid-Open No. 2005-012248 A.Kano, K.Doi, H.MacMahon, D.Hassell, MLGiger, “Digital image subtraction of temporally sequential chest images for detection of interval change”, (USA), Medical Physics, AAPM, Vol.21, Issue 3, March 1994, p.453-461

  However, in the method described in Patent Document 5, an optimal combination is selected for each image unit. However, in practice, the optimal combination may not be obtained at the local region level in the image. There is room for further improvement in terms of improvement in accuracy and reduction of artifacts in the difference image.

  The present invention has been made in view of the above circumstances, and aligns a subject in an image between a first image group composed of two or more images representing the same subject and a second image. A method and an apparatus for generating a difference image with higher accuracy regardless of a change in posture or state of the subject when generating a difference image between the first image group and the second image after registration The purpose is to provide a program.

  According to the image processing method of the present invention, the position of the subject in each image of the first image group composed of two or more images representing the same subject is set to the second one representing the same subject as the first image group. Alignment processing is performed to match the position of the subject included in the image, and for each pixel of each image in the first image group after the alignment processing, the difference in pixel value from the corresponding pixel in the second image And, for each corresponding pixel of each image in the first image group after the alignment process, the calculated difference is weighted so that the contribution becomes smaller as the absolute value of the difference increases. Thus, a composite difference image representing a difference between the first image group and the second image is generated.

  The image processing apparatus according to the present invention implements this method. That is, the subject included in the second image representing the same subject as the first image group is the position of the subject in each image of the first image group composed of two or more images representing the same subject. Positioning means for matching the position of the first image group, difference calculation means for obtaining a pixel value difference with a corresponding pixel in the second image for each pixel of each image of the first image group after the positioning process; The obtained difference is weighted and synthesized so that the contribution becomes smaller as the absolute value of the difference is larger for each corresponding pixel of each image in the first image group after the alignment process, There is provided a composite difference image generation means for generating a composite difference image representing a difference between one image group and the second image.

  Furthermore, an image processing program according to the present invention causes a computer to perform the above method (functions as each of the above means). That is, the computer includes the position of the subject in each image of the first image group composed of two or more images representing the same subject in the second image representing the same subject as the first image group. Difference calculation for obtaining a difference between a pixel value of a corresponding pixel in the second image for each pixel of each image in the first image group after the alignment processing, The means and the calculated difference are weighted for each corresponding pixel of each image in the first image group after the alignment process so that the contribution is reduced as the absolute value of the difference is increased. To function as a composite difference image generation means for generating a composite difference image representing a difference between the first image group and the second image.

  Next, details of the present invention will be described.

  As an application example of the image processing of the present invention, a medical image representing the chest of a human body is a processing target, a first image group is a past image captured in the past, a second image is a past image as a current image to be interpreted, It is conceivable to generate a differential image representing the change in the chest over time from the time when the image is taken to the present time, and to use it for the diagnosis of the presence or absence of an abnormal shadow representing a lesion in the image.

  “Alignment process” is a process of paying attention to the position of the subject in the two images to be processed, obtaining the correspondence between the positions of each pixel, and obtaining the shift amount from one image to the other image for each pixel. It is preferable that As a specific example, a rough alignment process (linear displacement is performed by detecting an overall shift amount (position shift amount) between two images by the temporal subtraction technique described in the above-mentioned patent document and non-patent document ( After performing global matching, local shift processing (local matching) is performed to detect the shift amount for each local region in both images, and to determine the shift amount of each pixel based on the detected shift amount. A method of performing alignment by performing non-linear distortion conversion processing (warping) on one image based on the shift amount is considered. Note that the alignment method may be performed by combining such linear and nonlinear alignment, or may perform only one of linear or nonlinear alignment.

  In the process of “combining the differences by weighting so that the contribution becomes smaller as the absolute value of the difference is larger” for each corresponding pixel of each image in the first image group after the alignment process, The weighting may be performed continuously according to the absolute value or may be a discontinuous weighting. As an example of discontinuous weighting, the minimum value of the absolute value of the difference is the pixel value of the corresponding pixel of the composite difference image, and the other difference values are weighted to 0, or the absolute value of the difference Is greater than a predetermined threshold, the weight is set to 0, and differences whose absolute value is less than or equal to the threshold are weighted equally, that is, the average value of the differences whose absolute value is less than or equal to the threshold is obtained, The pixel value of the corresponding pixel can be considered.

  Further, for each image in the first image group after the alignment processing, the dispersion degree of the absolute value of the difference in each pixel in the image is obtained, and an image that is small enough to satisfy the predetermined standard is obtained. You may make it produce | generate and produce | generate the synthetic | combination difference image showing the difference of the selected image and a 2nd image. Here, the “dispersion degree” is the degree of dispersion of the absolute value of the difference, and examples of specific index values include dispersion and standard deviation. In addition, as a specific example of “selecting an image whose degree of dispersion satisfies a predetermined criterion”, an image whose degree of dispersion is smaller than a predetermined threshold is selected, or a predetermined number of images in order of increasing degree of dispersion. Can be considered. When there is a single image selected in this way, an image based on the difference between the selected image and the second image is a composite difference image, and when there are a plurality of images selected in this way. For example, for each corresponding pixel in the selected image, a composite difference image may be generated or selected by using the minimum value of the absolute value of the difference as the pixel value of the corresponding pixel of the composite difference image. For each corresponding pixel in the image, the composite difference image may be generated using the average value of the differences as the pixel value of the corresponding pixel in the composite difference image. In addition, for each corresponding pixel of each selected image, if the weighting is combined so that the contribution becomes smaller as the absolute value of the difference is larger, how is the corresponding pixel of the combined difference image? The pixel value may be determined. Note that when the position of the subject in the image is completely matched by the alignment processing, the scatter degree of this difference is 0, whereas the greater the scatter degree of this difference is, the larger the difference is. This means that small parts are mixed and the accuracy of alignment is low. Therefore, if an image that is small enough to satisfy the predetermined standard is selected and a composite difference image is generated based on the selected image, a portion having a large difference is excluded and a composite difference image is generated. This means the same thing as “combining weights so that the contribution becomes smaller as the absolute value of the difference is larger”.

  In the image processing method, apparatus, and program of the present invention, alignment processing is performed between each of the first image group including two or more images representing the same subject and the second image, and after the alignment processing For each pixel of each image in the first image group, a difference in pixel value from the corresponding pixel in the second image is obtained, and the obtained difference is determined for each of the first image group after the alignment processing. For each corresponding pixel of the image of, a composite difference image representing the difference between the first image group and the second image is synthesized by weighting and combining so that the contribution becomes smaller as the absolute value of the difference is larger. Generate.

FIG. 1 schematically shows the effect of this process. FIGS. 1 (a) represents the pixel value at each position two images P of the first image group _n1, P _n2 and the direction of the second image P _A. 1 (b) is the difference between the images P _n1 of the second image P _A in the first image group at each position of the above, and the second image P _A in the first image group in the image The pixel value of the difference from P _n2 is represented. Here, it is assumed that abnormal shadow representing a lesion only the second image P _A is present, between the second image P _A in the first image P _n2 of images, the alignment processing results, the position It is assumed that there is a deviation and an image due to the difference between the two images appears as an artifact.

In FIG. 1B, the contribution of the difference pixel value becomes smaller as the absolute value of the difference is larger for each corresponding pixel of each of the images P _n1 and P _n2 of the first image group after the alignment processing. If the image is weighted and synthesized, the artifact portion in the difference from the image P _n2 having a large absolute value of the difference has a small contribution to the pixel value of the synthesized difference image, and the artifact in the synthesized difference image is reduced. The

On the other hand, among the portions where no positional deviation occurs between the image P _A and the image P _n2 , there is a portion with higher alignment accuracy than between the image P _A and the image P _n1. In this case, weighting is performed so that the contribution becomes smaller as the absolute value of the difference is larger, whereby the pixel value of the combined difference image of the pixel value based on the difference between the image P _A and the image P _n2 with higher alignment accuracy is obtained. The contribution of increases. In this way, the accuracy of the combined difference image is improved by generating the combined difference image so that the alignment accuracy is increased for each local region or pixel of each image in the first image group.

Note that the abnormal shadow portion of the image P _A has a large difference from both of the images P _n1 and P _n2 , and therefore, even if weighted so that the contribution becomes smaller as the absolute value of the difference increases, The composite difference image is not greatly reduced. Furthermore, in practice, the difference pixel value of the abnormal shadow portion does not become larger than the difference pixel value due to the artifact due to the positional deviation of the structure of the subject. Therefore, by appropriately setting the weighting rule, the abnormal shadow portion It is also possible not to reduce the pixel value of the difference.

  As described above, by performing image processing according to the present invention, it is possible to reduce artifacts due to differences in the posture of the subject for each radiographing and changes in the state, and to generate a more accurate differential image. Contributes to improved accuracy and diagnostic efficiency.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 2 shows an outline of the configuration of the chest image diagnosis support system according to the embodiment of the present invention. As shown in the figure, an image capturing / reading system 20 and an image management system 30, an image processing system 10, and an image display system 40 are connected in a communicable state via a network 50 such as a LAN.

  The image photographing / reading system 20 acquires a radiographic image obtained by photographing a patient's chest, and a CR (Computed Radiography) device 21 or an X-ray image digital photographing device (hereinafter referred to as an FPD device) 22 equipped with an FPD. Etc. are included.

  The image processing system 10 performs image processing on the radiographic image captured by the image capturing / reading system 20 to generate an image suitable for interpretation by a diagnostician, and includes an image processing server 11 and the like. The image processing method (apparatus) of the present invention is mounted on this image processing server 11.

  The image management system 30 stores and manages images generated by the image capturing / reading system 20 and the image processing system 10, and includes an image management server 31, a large-capacity external storage device 32, and database management software (for example, ORDB). (Object Relational DataBase) management software). Each image is associated with subject information such as subject ID, subject name, gender, date of birth, and examination information such as examination date and time, examination site, examination result, etc., and is managed by database management software Is stored in the large-capacity external storage device 32.

  The image display system 40 displays an image generated by the image capturing / reading system 20 or the image processing system 10, and includes a client PC 41, a high-definition liquid crystal display 42, and the like. It is also possible to set interpretation conditions such as an image to be interpreted and an image processing method for the image using an input device such as a keyboard and a mouse of the client PC 41.

An image processing method (apparatus) according to an embodiment of the present invention includes a past image group including two or more medical images (past images) P ₁ to P _N representing a chest of a subject that have been photographed in the past. image, the interpreted performs alignment between the image (image to be interpreted) P _a representing the chest of the same subject in the past image group, the difference between the past image group and the image to be interpreted after the positioning The function which produces _| generates the synthetic | combination difference image Qc showing is realized.

FIG. 3 is a block diagram showing a logical configuration and a data flow of the image processing server 11 in which the image processing function according to the embodiment of the present invention is implemented. As shown, the image management system 30, from the past image P ₁ of each of the past images and acquiring means 1 for acquiring and P _N and interpretation target image P _A, each of the past images of the image P _n ( n = 1, 2,..., N) The position of the chest in the interpretation target image PA and the position adjustment means 2 for matching the position of the chest in the interpretation target image P _A and each image of the past image group after the position alignment processing P _n pixels "(x, y) (x = 1,2, ···, X, y = 1,2, ···, Y) for each of the corresponding pixel in the image to be interpreted P _a difference △ _n (x, y) and the difference calculation means 3 for determining the, the obtained difference △ _n (x, y) and, for each corresponding pixel of the image P _n of each of the past images after the positioning process, Table by contributing as the absolute value is large synthesized weighted so as to reduce the difference (from P ₁ P _n) past images and the image to be interpreted P _a difference △ _n (x, y) Synthesis difference image generation unit 4 for generating an be combined difference image Q _c, a transmitting unit 5 for transmitting and synthesizing the difference image Q _c and interpretation target image P _A to the client PC 41.

  Each of the above means executes the program installed in the image processing server 11 from a storage medium such as a CD-ROM, thereby executing the CPU, main storage device, external storage device, input / output interface, operating system of the image processing server 11 Realized in cooperation with the system. In addition, the processing order by each of the above means is controlled by this program.

  Next, the flow of processing performed in the embodiment of the present invention will be described.

First, in the image capturing / reading system 20, at the time of examination, for example, the patient's chest is imaged by the FPD device 23, and an image P _n (n = 1, 2,..., N, A) is generated. Is done. Each of the generated images is associated with accompanying information such as subject information and examination information, and each image is stored in a file and output. The output image file is transmitted to the image management system 30 via the network 50.

  In the image management system 30, the image management server 31 receives the transmitted image file, and based on the data format and data structure defined by the database management software, the image data in the received image file is used as accompanying information. The data is stored in the large capacity external storage device 32 in association with each other. The stored image data is in a searchable state using part or all of the accompanying information such as the subject ID and the examination date and time as a search key.

The interpreter designates accompanying information such as patient ID and photographing date and time and an interpretation pattern (comparison interpretation based on a difference from a past image) and requests execution of the process by operating the client PC 41. Here, P _N and the image to be interpreted P _A from the image P ₁ of the past image group by specifying the associated information described above are identified, the contents of the image processing necessary for the image interpretation by specifying the interpretation pattern specific Is done.

  The accompanying information and interpretation pattern specified by the client PC 41 are transmitted to the image processing server 11, and the image processing server 11 generates an image necessary for interpretation based on the received interpretation pattern, that is, an image processing program. A program for causing the image processing server 11 to function as each unit in FIG. 3 is started.

  FIG. 4 is a flowchart showing the flow of processing executed by the activated image processing program.

First, after n = 0 is set by initialization processing of control parameters used in the program (step a1), the acquisition unit 1 of the image management system 30 based on the search condition based on the accompanying information received from the client PC 41 is used. Send a search request for the database. In the image management server 31 performs a database search in response to the received search request, retrieves the P _N, P _A from the image P ₁ that matches the search condition, and transmits to the image processing server 11. The acquisition unit 1 of the image processing server 11 receives the images P ₁ to P _N and P _A and temporarily stores them in the main storage device or the external storage device of the image processing server 11 (step a2).

Then, the value of subscript n identifies the image is incremented by one, after becoming a n = 1 (step a3), the alignment means 2, image P _n (= P ₁₎ and reads the P _A, the image P the position of the chest in ₁ performs alignment processing to align the position of the breast contained in the image to be interpreted P _a, generates an image P ₁ "after the positioning process (step a4). here, the aforementioned patents The temporal subtraction method described in Document 1 etc. is used, which will be described in detail below.

First, schematic alignment processing between the image P ₁ and the image P _A (global matching). This image P ₁ so as to coincide with the image P _A, affine transformation (rotation, translation shift) with respect to P ₁ is a process for performing an image P ₁ by this process, P as shown in FIG. 5 Converted to ₁ '.

  After the global matching process is completed, a local alignment process (local matching) is performed.

Specifically, as shown in FIG. 6, the image P _A is divided into horizontal I × longitudinal J local areas R _A [1,1] to R _A [I, J], and the central pixel of each local area is divided. Are represented by an xy coordinate system (x, y). Also, search ROIR ₁ '[1, 1], ..., R ₁ ' [i, j], ..., R ₁ '[I, J] are set in the image P ₁ '. This search ROI is set corresponding to each local area R _A [1,1] to R _A [I, J] of the image P _A and has the same center coordinates (x, y). The region is larger than R _A [1,1] to R _A [I, J]. Here, it is assumed that each local region R _A [1,1] to R _A [I, J] is a region that is four times as large (twice both vertically and horizontally).

Among the search ROIs set in the image P ₁ ′, _{RA A} [I, J] is moved from each local region R _A [1,1] of the image P _A , and the matching degree of the local region is the highest. The position (center position (x ′, y ′) of R _A [I, J] from each local area R _A [1,1]) is obtained for each search ROI. As the index value indicating the level of matching, an index value based on the least square method or cross-correlation can be used.

_A shift value (Δx, Δy) for each central pixel (x, y) of each local region R _A [1,1] to R _A [I, J] (where Δx = x′−x, Δy = y). '-Y) is obtained. FIG. 7 schematically shows the central pixels (x, y) and (x ′, y ′) of each local region of the images P _A and P ₁ ′, and the displacement Δ ₁ [1,1] between the images. It is a thing. Further, a two-dimensional tenth-order fitting polynomial is used to obtain shift values (Δx, Δy) for all the pixels of the image P ₁ ′ using the shift values (Δx, Δy) for each central pixel (x, y). performs approximation processing by, based on a shift value for each pixel obtained ([Delta] x, [Delta] y), the image P ₁ 'each pixel of the (x, y) nonlinear distortion conversion process of shifting the (warping) of the image P _1' To generate an image P ₁ ″ after alignment.

When the alignment processing between the images P ₁ and P _A is completed as described above, the difference calculating unit 3 calculates the difference between the pixel (x, y) of the image P ₁ ″ and the corresponding pixel in the image P _A. Δ ₁ (x, y) is obtained (step a5).

Then, the value of subscript n identifies the image is incremented by one, after becoming a n = 2 (step a6, a3), in the same manner as described above, the position adjustment process of images P ₂ and P _A by the alignment means 2 (step a4), the calculation of the difference △ ₂ image P ₂ "and P _a by the difference calculating means (x, y) (step a5) is carried out, further, is repeated until the same process is n = n ( Step a6) Finally, Δ _N (x, y) is obtained from the difference Δ ₁ (x, y).

The composite difference image generation means 4 obtains a composite difference Δ _c (x, y) based on each difference Δ ₁ (x, y) to Δ _N (x, y) (step a7). Specifically, for each corresponding position (coordinate) of each difference, the one having the smallest absolute value of the difference is selected and set as the combined difference Δ _c (x, y).

Further, the synthesized difference image generating means 4 generates a synthesized difference image Q _c based on the synthesized difference Δ _c (x, y) (step a8). Specifically, the pixel value Q _c (x, y) of each pixel of the composite difference image Q _c is obtained by the following equation (1).
Q _c (x, y) = cont · Δ _c (x, y) + mid (1)
Where cont is a coefficient that expresses the degree of difference enhancement
mid: Intermediate value of the number of gradations of the image (mid = 512 for 10-bit images)
Transmission means 5 transmits the image data of the generated synthesized difference image Q _c and interpretation target image P _A to the client PC 41 (step a9).

Client PC41 displays the synthesized difference image Q _c and interpretation target image P _A on the basis of the image data received from the image processing server 11 to the high-resolution liquid crystal display 42. FIGS. 8C and 8B schematically show displayed images.

As described above, in the embodiment of the present invention, in the image processing server 11, the alignment means 2, the alignment processing between the image P ₁ of each of the past image group and P _N and the image to be interpreted P _A performed, the difference calculating unit 3, for each pixel of the "P _N from" image P ₁ of each of the past images after the positioning process, the difference between the corresponding pixel in the image to be interpreted P _a △ ₁ (x, y) is obtained from Δ _N (x, y), and the composite difference image generating means 4 determines the minimum absolute value of the differences Δ ₁ (x, y) to Δ _N (x, y) for each pixel. _c (x, y), and a synthesized difference image Q _C is generated based on this synthesized difference Δ _c (x, y). Here, the composite difference image generation means 4 uses the minimum absolute value of the differences Δ ₁ (x, y) to Δ _N (x, y) as the composite difference Δ _c (x, y). In determining the pixel value of each pixel of the image Q _C , only the smallest absolute value of the difference Δ ₁ (x, y) to Δ _N (x, y) contributes, and the others do not contribute at all. Will not be. That is, from the images P ₁ for each pixel of the P _N in the past images, select the one with the highest positioning accuracy of the image to be interpreted P _A, it means that generates a composite difference image Q _C. FIG. 8 schematically shows this. As shown, in the image processing according to the present invention, for each pixel of each image of the past image group, the image P _n generated by selecting with the highest positioning accuracy of the image to be interpreted P _A "( 8 (a)) and the image Q _C resulting from the difference between the image to be interpreted P _A. Therefore, the difference in posture of the subject and the artifact due to the change in state are reduced, and the difference is more accurate. An image can be generated, which contributes to improvement of diagnosis accuracy and diagnosis efficiency.

In the above description, the positioning process and the difference calculation processing between the image P ₁ of each of the past image group and P _N and the image to be interpreted P _A, it was performed sequentially for each image of the past image group However, as shown in FIG. 9, it may be performed by parallel processing for each image in the past image group. That is, after acquiring P _N and P _A from the image P ₁ (step b1), the difference Δ ₁ (x, y) is calculated from the alignment (step b2-1) between the images P ₁ and P _A (step b3). -1) and processing from the alignment of the images P ₂ and P _A (step b2-2) to the calculation of the difference Δ ₂ (x, y) (step b3-2).・ Processing from registration (step b2-n) of images P _n and P _A to calculation of difference Δ _n (x, y) (step b3-n),..., And images P _N and P _A From the alignment (step b2-N) to the calculation of the difference Δ _n (x, y) (step b3-N) in parallel, the difference Δ ₁ (x, y) to Δ _n (x, y) are calculated at the same time, and based on these, a composite difference Δ _C (x, y) is calculated (step b4), a composite difference image Q _C is generated (step b5), and images Q _C and P _A are transmitted. (Su Perform step b6). Processing time can be reduced by this parallel processing.

In the above description, the alignment process and the difference calculation are performed for each image in the past image group. However, the difference calculation and the composite difference calculation may be performed for each corresponding pixel. 10 and 11 are flowcharts showing the flow of processing in this case. As shown in FIG, and the parameter n, m, x, and 0 to initialize the y (step c1), after the acquisition unit 1 acquires the P _N and P _A from the image P ₁ (step c2), the image P aligns _n and P _a, to obtain a "P _n from" image P ₁ after the positioning While (from step c3 c5) until it is the same as described above, then the position (1,1) ( Steps c6 and c7), sequentially from the difference Δ ₁ (1,1) between the pixel values of the image P ₁ ″ and the image P _A to the difference Δ _N (1,1) between the pixel values of the image P _N ″ and the image P _A Calculate (steps c8 to c10), and based on the calculated differences Δ ₁ (1,1) to Δ _N (1,1), calculate the composite difference Δ _C (1,1) at the position (1,1). (Step c12), and further, positions (2, 1), ..., (X, 1), (1, 2), ..., (X, 2), ..., (1, Y) , ... , (X, Y) for (step c6, c7, c13, c14, c15), " the difference △ ₁ of the pixel values of the image P _A (x, y) from the image P _N" image P ₁ and the image P _A The pixel value difference Δ _N (x, y) is sequentially calculated (steps c8 to c10), and the position (x, y) is calculated based on the calculated difference Δ ₁ (x, y) to Δ _N (x, y). The composite difference Δ _C (x, y) in y) is calculated (step c12). Then, the composition difference at all positions △ _C (x, y) based on the generated composite difference image Q _C (Step c16), and transmits together with the image P _A to the client PC 41 (Step c17).

Furthermore, as shown in FIG. 12, from (step d3-1 after conducted in parallel from the image P ₁ of each of the past images of the position adjustment process of the P _N and the image to be interpreted P _A for each image d3- N), the calculation of Δ _N (x, y) from the difference Δ ₁ (x, y) between each image P ₁ to P _N and the image P _A and the calculation of the composite difference Δ _C (x, y). It may be performed in parallel for each pixel (steps d4-1 to d7-1, d4-2 to d7-2,..., D4-K to d7-K; K = X · Y, Represents the total number of pixels in the image). Thus, the composite difference Δ _C (1,1),..., Δ _C (X, 1), Δ _C (1,2),..., Δ _C (X, 2),. _{Since C} (1, Y),..., Δ _C (X, Y) are calculated at the same time, the processing time can be shortened.

Further, as shown in the block diagram of FIG. 13, synthesized the difference image generation unit 4, the difference △ ₁ from the images P ₁ of the past images calculated by the difference calculation means 3 and P _N and the image to be interpreted P _A (x, y) on the basis of the △ on _n (x, y), the image units of previous images, determine the degree of dispersion V _n of the absolute value of the difference, 3 images from those spread of V _n is small (P _n1 , P _n2, P _n3) a selection unit 4a for selecting, selected 3 images (P _n1, P _n2, the difference between P _n3) and the image _{P a △ n1 (x, y} ), △ n2 (x, y), △ _n3 (x, based on y) synthesis difference △ _C (x, a synthetic difference calculation unit 4b for calculating a y), the calculated synthesized difference △ _C (x, synthetic differential image based on y) it may be composed of a synthetic difference image generation unit 4c for generating a Q _C.

Here, the selection unit 4a first determines the difference in the pixels in the lung field region among Δ _n (x, y) (n = 1, 2,..., N) in units of images of the past image group. The dispersion degree V _n of the absolute value of the difference is obtained. Specifically, the variance of the absolute value of the difference is calculated based on the following equation (2).

Note that a known method (for example, a method described in Japanese Patent Laid-Open No. 2003-006661) can be used as a method for detecting a lung field region in an image.

Further, the selection unit 4a among the calculated variance V ₁ of the V _N, select the three in order of smaller value, selects the past image P _n1, P _n2, P _n3 corresponding to them. Note that the number of images to be selected is not limited to three. Alternatively, it is acceptable to set the predetermined threshold value Th _V of the dispersion V _n in advance, may be selected past image corresponding to the intended dispersion V _n is smaller than the threshold value Th _V.

Then, the composite difference calculation unit 4b is based on the differences Δn ₁ (x, y), Δ _n2 (x, y), Δ _n3 (x, y) of the selected past images P _n1 , P _n2 , P _n3. Then, the average value of the differences is calculated as a composite difference Δ _c (x, y) for each position corresponding to each difference, and the composite difference image generation unit 4 c is based on the composite difference Δ _c (x, y). The composite difference image Q _c is generated based on the equation (1). The composite difference calculation unit 4b calculates the minimum absolute value of the differences Δn ₁ (x, y), Δ _n2 (x, y), Δ _n3 (x, y) for each position corresponding to each difference. The composite difference Δ _c (x, y) may be used.

Thus, selection portion 4a, by the degree of scattering V _n of the absolute value of the difference calculated by the image units of the past image group selects the small image, the large image dispersion degree V _n of synthetic difference image Q _C It is no longer used for generation. Here, an image with a high degree of dispersion V _n that has a high possibility of artifacts due to misalignment in the alignment process includes a portion with a large absolute value of the difference Δ _n (x, y). Therefore, the difference △ _n (x, y) pixels having large absolute value of not contribute to the determination of the pixel value of the composite difference image Q _C, the same effect as the above embodiment can be obtained.

Moreover, since the synthetic | combination difference calculation part 4b calculates the average value of a difference for every position corresponding to each difference (DELTA) _n (x, y) as synthetic | combination difference (DELTA) _c (x, y), difference (DELTA) _n (x, y) ), The discontinuity of the pixel values between the pixels in the image is reduced, and the composite difference image Q _C is easier to see compared to the case where the absolute value of) is the minimum as the composite difference Δ _c (x, y). It becomes an image.

In the above embodiment, the difference △ _n (x, y) based on the composite difference △ _C (x, y) had to calculate the, by the following equation (1) ', the images P _n and P _a A difference image Q _n representing a difference is generated, and based on a difference Δ _Qn (x, y) between a pixel value Q _n (x, y) of each pixel in the difference image Q _n and an intermediate value of the number of gradations of the image. , it may be determined a pixel value of each pixel of the composite difference image Q _C. For example, it is conceivable that the pixel value at the corresponding position in the composite difference image Q _{C is} the one having the smallest absolute value of Δ _Qn (x, y) for each corresponding position in each difference image Q _n .
Q _n (x, y) = cont · Δ _n (x, y) + mid (1) ′
In the above embodiment, the system configuration is such that the processing is distributed between the image processing server 11 and the client PC 41 connected via the network 50. However, these processing are performed by a single computer such as a diagnostic workstation. Also good.

The figure which represented typically the effect of this invention based on distribution of the pixel value of two past images and an image to be interpreted The figure which showed the outline | summary of the structure of the chest image-diagnosis assistance system used as embodiment of this invention. The block diagram showing the logical structure and data flow of the image processing server by which the image processing function by this invention is mounted The flowchart showing the flow of image processing according to the present invention. The figure explaining the global matching process which a positioning means performs The figure explaining the local matching process which an alignment means performs A diagram schematically showing the shift of the center pixel of each local area obtained by local matching processing performed by the positioning means The figure which represented typically the relationship between the past image group, the interpretation target image, and a synthetic | combination difference image in embodiment of this invention. The flowchart showing another example of the flow of the image processing of this invention The flowchart showing the other example of the flow of image processing of the present invention (the first half) Flowchart showing another example of the flow of image processing according to the present invention (second half) The flowchart showing another example of the flow of the image processing of this invention The block diagram showing the detailed structure of the synthetic | combination difference image generation means in other embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Acquisition means 2 Positioning means 3 Difference calculation means 4 Composite difference image generation means 4a Selection part 4b Composite difference calculation part 4c Composite difference image generation part 5 Transmission means 10 Image processing system 11 Image processing server 20 Image photographing / reading system 21 CR Device 22 X-ray digital imaging device equipped with FPD 30 Image management system 31 Image management server 32 Large capacity external storage device 40 Image display system 41 Client PC
42 High-definition LCD 50 Network

Claims (9)

Alignment for aligning the position of the subject in each image of the first image group consisting of two or more images representing the same subject with the position of the subject included in the second image representing the subject Process
For each pixel of each image in the first image group after the alignment process, obtain a difference in pixel value from the corresponding pixel in the second image,
The difference is weighted for each corresponding pixel of each image in the first image group after the alignment process so that the contribution becomes smaller as the absolute value of the difference is larger, An image processing method, comprising: generating a composite difference image representing a difference between a first image group and the second image.   The minimum value of the absolute value of the difference is set as the pixel value of the corresponding pixel of the composite difference image for each corresponding pixel of each image of the first image group after the alignment process. Item 8. The image processing method according to Item 1. For each image of the first image group after the alignment process, obtain the degree of dispersion of the absolute value of the difference at each pixel in the image,
Select an image that is so small that the degree of dispersion meets a predetermined standard,
The image processing method according to claim 1, wherein the composite difference image representing a difference between the selected image and the second image is generated. Alignment for aligning the position of the subject in each image of the first image group consisting of two or more images representing the same subject with the position of the subject included in the second image representing the subject Means,
Difference calculating means for obtaining a pixel value difference with a corresponding pixel in the second image for each pixel of each image of the first image group after the alignment;
The difference is weighted for each corresponding pixel of each image in the first image group after the alignment process so that the contribution becomes smaller as the absolute value of the difference is larger, An image processing apparatus comprising: a synthesized difference image generating unit that generates a synthesized difference image representing a difference between a first image group and the second image.   For each corresponding pixel of each image of the first image group after the alignment processing, the combined difference image generation unit calculates the minimum value of the absolute value of the difference as the pixel value of the corresponding pixel of the combined difference image The image processing apparatus according to claim 4, wherein:   The composite difference image generation means obtains a dispersion degree of the absolute value of the difference in each pixel in the image for each image of the first image group after the alignment process, and the dispersion degree is a predetermined degree. 6. The image processing according to claim 4, wherein an image that is small enough to satisfy a criterion is selected, and a composite difference image that represents a difference between the selected image and the second image is generated. apparatus. Computer
Alignment for aligning the position of the subject in each image of the first image group consisting of two or more images representing the same subject with the position of the subject included in the second image representing the subject Means,
Difference calculating means for obtaining a pixel value difference with a corresponding pixel in the second image for each pixel of each image of the first image group after the alignment;
The difference is weighted for each corresponding pixel of each image in the first image group after the alignment process so that the contribution becomes smaller as the absolute value of the difference is larger, An image processing program that functions as a composite difference image generation unit that generates a composite difference image representing a difference between a first image group and the second image. The computer,
For each corresponding pixel of each image of the first image group after the alignment processing, the combined difference image generation unit calculates the minimum value of the absolute value of the difference as the pixel value of the corresponding pixel of the combined difference image The image processing program according to claim 7, wherein the image processing program is made to function as follows. The computer,
The composite difference image generation means obtains a dispersion degree of the absolute value of the difference in each pixel in the image for each image of the first image group after the alignment process, and the dispersion degree is a predetermined degree. The image according to claim 7 or 8, wherein an image that is small enough to satisfy a criterion is selected, and a function is generated to generate a composite difference image that represents a difference between the selected image and the second image. Processing program.

Images