JP6910205B2 - Disease discrimination support device and computer program - Google Patents

Description

The present invention relates to a technique for supporting the differentiation of a plurality of diseases.

There are many types of dementia such as Alzheimer's disease and Lewy body dementias, and the number of patients is the largest in the order of Alzheimer's disease (AD) and Lewy body dementias (DLB), but both have similar symptoms. There is.

Conventionally, by statistically processing brain images of a plurality of patients, a disease-specific region of interest showing a decrease in blood flow specific to a disease is created, and this is applied to the brain of a subject to apply the region of disease-specific interest. Diagnosis is made based on the blood flow condition in.

For example, a disease-specific region of interest in AD is created, and the index of decreased blood flow in the region of interest of the subject is analyzed to observe the severity and spread of the disease.

However, it was not possible to distinguish whether the subject was AD or DLB using the disease-specific region of interest in AD.

Therefore, an object of the present invention is to provide a disease-specific region of interest useful for differentiating a plurality of diseases, and to provide an index showing high discrimination ability based on the region of interest.

A computer program according to one aspect of the present invention is a first data relating to images of a predetermined site of a large number of healthy subjects and data relating to images of the predetermined site of a large number of patients having a first disease. It is a computer program for causing a disease discrimination support device having a storage unit to store patient data and second patient data which is data related to images of a plurality of patients having a second disease at the predetermined site. , The healthy subject data is compared with the first patient data to create a first disease data relating to the first disease at the predetermined site, and the healthy subject data is compared with the second patient data. Then, the step of creating the second disease data related to the second disease at the predetermined site and the first disease data are corrected based on the second disease data, and the disease related to the first specific disease is corrected. The step of creating the first region data indicating the specific region of interest and creating the second region data indicating the disease-specific region of interest related to the second specific disease based on the second disease data, and the disease discrimination. Let the support device do it.

In a preferred embodiment, the first disease data has data indicating a disease-specific region relating to the first disease, and the second disease data has data indicating a disease-specific region relating to the second disease. Then, in the step of creating the first region data, from the disease-specific region of the first disease data, the region where the disease-specific region of the first disease data and the disease-specific region of the second disease data overlap. The remaining area excluding the above may be created as the first area data.

In a preferred embodiment, the disease-specific region of interest of the first region data and the disease-specific region of interest of the second region data may partially touch each other without overlapping each other.

In a preferred embodiment, in the step of creating the second disease data, the disease group consisting of the second patient data and the healthy subject group consisting of the healthy subject data are compared between the groups, and statistically statistically for each pixel. The second disease data is obtained by performing a test and combining the 2-1 disease data showing the disease-specific region based on the first significance level and the 2-2 disease data showing the disease-specific region based on the second significance level. In the step of creating the first region data and the second region data, from the disease-specific region of the first disease data to the disease-specific region of the first disease data and the 2-1 disease. The remaining region excluding the region overlapping the disease-specific region of the data may be created as the first region data, and the disease-specific region of the 2-2 region data may be used as the second region data.

In a preferred embodiment, in the step of creating the second region data, from the disease-specific region of the second disease data to the disease-specific region of the first disease data and the disease-specific region of the second disease data. The remaining area excluding the area overlapping with may be created as the second area data.

A computer program according to another aspect of the present invention includes subject data, which is data related to an image of a predetermined part of a subject, healthy person data, which is data related to an image of a predetermined part of a large number of healthy persons, and claims 1 to 5. A computer program to be executed by a disease discrimination support device having a storage unit for storing the first region data and the second region data created by the computer program according to any one of the above, wherein the subject data and the subject data are described. The step of comparing the data of healthy subjects to calculate the feature amount, and the disease-specific interest of the first feature amount and the second area data calculated based on the feature amount related to the disease-specific interest region of the first region data. The disease discrimination support device is made to execute a step of calculating an index based on the second feature amount calculated based on the feature amount of the region and a step of outputting the index.

In a preferred embodiment, the feature amount may be a Z value calculated based on the subject data and the healthy person data.

In a preferred embodiment, the image of the predetermined site may be a SPECT (Single Photon Emission Tomography) tomography image of the brain or a PET (Positron Emission Tomography) tomography image.

In a preferred embodiment, the first disease may be Alzheimer's disease and the second disease may be Lewy body dementias.

The disease discrimination support device according to one aspect of the present invention is data relating to images of a predetermined site of a large number of healthy subjects and data relating to images of the predetermined site of a large number of patients having a first disease. A storage unit that stores first patient data and second patient data that is data related to images of the predetermined sites of a plurality of patients having a second disease, the healthy person data, and the first patient data. To create first disease data relating to the first disease at the predetermined site, and compare the healthy subject data with the second patient data to relate to the second disease at the predetermined site. A disease-specific data creation means for creating a second disease data, and a first region in which the first disease data is corrected based on the second disease data to indicate a disease-specific region of interest related to the first specific disease. It is provided with a disease-specific region of interest creating means for creating data and creating second region data indicating a disease-specific region of interest related to the second specific disease based on the second disease data.

The disease discrimination support device according to another aspect of the present invention comprises subject data which is data related to an image of a predetermined part of a subject, data of a healthy person which is data related to an image of a predetermined part of a large number of healthy persons, and claim 10. A storage unit that stores the third region data and the fourth region data created by the disease discrimination support device, a feature quantity calculation means that compares the subject data and the healthy subject data, and calculates the feature quantity. The first feature amount calculated based on the feature amount of the disease-specific interest region of the first region data and the second feature amount calculated based on the feature amount of the disease-specific interest region of the second region data. An index calculation means for calculating an index based on the above and an output means for outputting the index are provided.

According to the present invention, it is possible to provide a disease-specific region of interest useful for differentiating a plurality of diseases, and to provide an index showing high discrimination ability based on the region of interest.

The overall block diagram of the disease discrimination support apparatus which concerns on embodiment of this invention is shown. It is a figure which shows an example of the data structure of the healthy person data and the standard deviation data. As a result of the statistical test, explanatory diagrams are shown for nine regions I to IX in which all voxels of patient data are classified as AD and DLB. An explanatory diagram for creating an AD area of interest is shown. A schematic diagram showing nine regions I to IX as brain regions is shown. It is a flowchart which shows the procedure which creates the disease-specific area of interest. It is a flowchart which shows the procedure which performs the identification of a subject.

Hereinafter, the disease discrimination support device and the computer program according to the embodiment of the present invention will be described with reference to the drawings. The disease discrimination support device and the computer program according to the present embodiment provide a disease-specific region of interest useful for differentiating a plurality of diseases, and provide an index showing high discrimination ability based on the region of interest. be.

FIG. 1 shows an overall configuration diagram of the disease discrimination support device 1 according to the present embodiment.

In the disease discrimination support device 1 according to the present embodiment, one or more of the disease discrimination support device main body 10, the input device 2, the output device 3, and the input / output device 4 connected to the disease discrimination support device main body 10 are connected. There is. The disease discrimination support device main body 10 is composed of, for example, a general-purpose computer system including a processor and a memory, and individual components or functions in the disease discrimination support device main body 10 described below include, for example, a computer program. It is realized by executing. This computer program can be stored on a computer-readable recording medium. The input device 2 may be, for example, a keyboard, a pointing device, or the like. The output device 3 may be, for example, a display device, a printer, or the like. The input / output device 4 may be, for example, a large-capacity storage device or a network interface device.

The disease discrimination support device main body 10 includes an AD patient data storage unit 11A, a DLB patient data storage unit 11B, a healthy person data storage unit 12, a normalization processing unit 13, a change region creation unit 14, and a disease-specific Area data storage unit 15, interest area creation unit 16, disease-specific interest area data storage unit 17, subject data storage unit 18, normalization processing unit 19, Z value calculation unit 20, Z value data storage unit It includes 21, an index calculation unit 22, a calculation data storage unit 23, a threshold data storage unit 24, and a discrimination unit (output unit) 25.

The AD patient data storage unit 11A captures the heads of a large number of patients with Alzheimer's disease (hereinafter referred to as AD (Alzheimer's Disease)) in dementia with a SPECT (Single Photon Emission Computed Tomography) imaging device. The voxel value for each pixel of the tomographic image and the number of samples are stored as AD patient data. In this embodiment, patient data of the heads of a large number of patients are stored as a plurality of diseases and for each of patients with Lewy body dementias (hereinafter referred to as DLB (Dementia with Lewy Bodies)). The DLB patient data storage unit 11B is a pixel-by-pixel boxel value and a number of samples of SPECT tomographic images of the heads of a large number of patients of Lewy body dementia in dementia (hereinafter referred to as DLB (Dementia with Lewy Bodies)). Is stored as DLB patient data. Further, the AD patient data and the DLB patient data may be acquired via, for example, the input device 2 or the input / output device 4, or may be acquired directly from the SPECT imaging device. In the present embodiment, the AD patient data corresponds to the first patient data, and the DLB patient data corresponds to the second patient data.

The healthy person data storage unit 12 stores the average value of the voxel values for each pixel of the tomographic images of the brains of a large number of healthy people and the number of samples collected in advance as healthy person data. Similarly, the standard deviation data is obtained by obtaining the standard deviation using the tomographic image data of the standard brain of a normalized healthy subject. The healthy person data may be acquired via, for example, the input device 2 or the input / output device 4.

The normalization processing unit 13 stores the voxel value of the AD patient data stored in the AD patient data storage unit 11A, the voxel value of the DLB patient data stored in the DLB patient data storage unit 11B, and the healthy person data storage unit 12. Normalize the voxel values of the stored healthy person data. Here, since the shape and size of the human brain vary from person to person, the voxel values are normalized to a predetermined shape and size, and the AD patient data and DLB patient data normalized to the same structure. And the healthy person data are compared. Further, the normalization processing unit 13 calculates the standard deviation of the voxel value for each pixel based on the normalized AD patient data, DLB patient data, and healthy person data. The data normalized in advance may be stored in the AD patient data storage unit 11A, the DLB patient data storage unit 11B, and the healthy person data storage unit 12, or the data after normalization by the normalization processing unit 13 may be stored. The data may be stored in the AD patient data storage unit 11A, the DLB patient data storage unit 11B, and the healthy person data storage unit 12, respectively.

FIG. 2 is a diagram showing an example of the data structure of the normalized healthy subject data and the standard deviation data. The healthy person data in FIG. 3A consists of N image data of XY cross sections in the Z-axis direction when the left-right direction of the head is the X-axis, the front-back direction is the Y-axis, and the up-down direction is the Z-axis. .. The voxel value included in each image data corresponds to the pixel value of each image. The standard deviation data in FIG. 3B has the same structure as the healthy person data, and stores the standard deviation obtained at the same time when the healthy person data, which is the average value of a large number of healthy person data, is calculated. There is.

The change region creation unit 14 creates a region in which blood flow is significantly lower than that of a healthy subject, based on the pixel data of the entire brain of AD patient data and DLB patient data. Specifically, the change region creation unit 14 includes a boxel value of AD patient data normalized by the normalization processing unit 13, a boxel value of DLB patient data, a standard deviation of AD patients and DLB patients, and AD patients and DLB patients. Statistical test (for example, t) for each pixel (boxel) based on the number of samples of the healthy person, the boxel value of the healthy person data normalized by the normalization processing unit 13, the standard deviation of the healthy person, and the number of samples of the healthy person. A test (one-sided test)) was performed, and when the p value (significance probability) <0.05 (or 0.01), it was assumed that there was a statistically significant difference, and the boxel judged to have a significant difference was significantly blood. The area where the flow is changing (decreasing). Then, the change region creation unit 14 creates disease-specific region data indicating a disease-specific region (AD region and DLB region) of each disease based on all regions where blood flow is reduced. In this way, a disease-specific region (VOI (Volume of Interest)) is created by comparing the disease group and the healthy subject group by a statistical test. The data in the AD region corresponds to the first disease data, the data in the DLB region corresponds to the second disease data, and the change region creation unit 14 corresponds to the disease-specific data creation means.

The disease-specific region data storage unit 15 stores the disease-specific region data created by the change region creation unit 14.

As a result of the above statistical test, all voxels of AD patient data and DLB patient data can be divided into nine regions for AD and DLB as a classification, as shown in FIG.

FIG. 3 shows explanatory diagrams for nine regions I to IX in which all voxels of AD patient data and DLB patient data are classified as AD and DLB as a result of statistical testing.

As shown in FIG. 3, AD has three classifications of "normal", "slightly decreased", and "significantly decreased", and DLB has three categories of "normal", "slightly decreased", and "significantly decreased". By multiplying the classification, it is divided into 9 regions I to IX. "Normal" and "slightly decreased" are the categories judged by statistical tests to be not significantly different. "Slightly decreased" indicates a category in which blood flow is not "significantly decreased" but is lower than "normal", and is not a category judged by a statistical test, but "normal". Since the boundary between is not clear, the boundary with "normal" is indicated by a dotted line. In addition, regions I to IX in the figure show the following contents.
I: Significantly decreased in DLB, normal region in AD II: Significantly decreased in DLB, slightly decreased in AD III: Significantly decreased in DLB, significantly decreased in AD IV: Slightly decreased in DLB, normal region in AD V: Slightly decreased in DLB, slightly decreased in AD VI: Slightly decreased in DLB, significantly decreased in AD VII: Normal in DLB, normal region in AD VIII: Normal in DLB, slightly reduced region in AD IX: Normal in DLB, significantly reduced region in AD

Then, as shown in FIGS. 4A and 4B, the AD region created by the change region creation unit 14 corresponds to regions III, VI, and IX, and the DLB region corresponds to regions I, II, and III. Equivalent to.

Further, when the regions I to IX are schematically represented as the region of the brain B, it is as shown in FIG.

FIG. 5 shows a schematic diagram showing nine regions I to IX as brain regions. As shown in FIG. 5, the AD region is represented as regions III, VI, IX and the DLB region is represented as regions I, II, III.

The region of interest creation unit 16 creates a region of disease-specific interest for AD and a region of disease-specific interest for DLB. Specifically, the region of interest creation unit 16 compares the AD region and the DLB region, which are disease-specific regions for each disease, stored in the disease-specific region data storage unit 15, and from the AD region to the AD region. The remaining area excluding the area where the DLB area and the DLB area overlap is created as an AD area of interest.

That is, as shown in FIGS. 4 (a), 4 (b), and 5, the AD region and the DLB region are regions in which the regions III overlap each other. Therefore, the region of interest creation unit 16 creates regions VI and IX obtained by removing region III from the AD region as an AD region of interest (AD-VOI). Further, the region of interest creation unit 16 does not process the DLB region, and sets the same regions I, II, and III as the DLB region of interest (DLB-VOI). In this way, the region of interest creation unit 16 creates disease-specific region of interest data indicating the region of disease-specific interest (AD region of interest, DLB region of interest) of each disease. The data in the AD area of interest corresponds to the data in the first area, and the data in the DLB area of interest corresponds to the data in the second area. Further, when the significance level of the DLB region for correcting the AD region and the DLB region serving as the DLB region of interest is the same, as shown in FIG. 5, the AD region of interest and the DLB region of interest do not overlap each other. It is the area where the parts touch. The region of interest creation unit 16 corresponds to the disease-specific region of interest creation means.

The disease-specific region of interest data storage unit 17 stores the disease-specific region of interest data created by the region of interest creation unit 16.

The subject data storage unit 18 stores the voxel value (hereinafter referred to as subject data) for each pixel of the SPECT tomographic image of the subject's head as data. Subject data may be acquired via, for example, the input device 2 or the input / output device 4.

The normalization processing unit 19 normalizes the subject data (voxel value data) of the subject stored in the subject data storage unit 18. Here, since the subject data after normalization is compared with the healthy subject data, the normalization processing unit 19 normalizes to the same structure as the healthy subject data. The subject data stored in advance may be stored in the subject data storage unit 18, or the subject data after normalization by the normalization processing unit 19 may be stored in the subject data storage unit 18.

The Z value calculation unit 20 calculates the Z value (feature amount) by comparing the corresponding pixels (pixels of the same portion) of the subject data and the healthy subject data. That is, the Z value calculation unit 20 calculates the boxel value of the subject data normalized by the normalization processing unit 19 and the boxel value (mean value and standard deviation) of the healthy subject data normalized by the normalization processing unit 13. Use to calculate the Z value. The Z value calculation unit 20 calculates the Z value for all valid pixels (that is, all voxel values) of the normalized subject data using the following formula.
Z value = {(Voxel value of healthy subject data)-(Voxel value of subject data)} / Standard deviation

The Z value data storage unit 21 stores the Z value calculated by the Z value calculation unit 20 as data.

The index calculation unit 22 calculates the total value of the Z values showing positive values for each disease in the disease-specific interest region (AD interest region, DLB interest region), and calculates the ratio thereof as an index (CIS (Cingulate Island Sign)). )). That is, the index calculation unit 22 has a positive value among the Z values stored in the Z value data storage unit 21 for each of the AD interest region and the DLB interest region stored in the disease-specific interest region data storage unit 17. The total value of the Z values (AD-SUM (first feature amount), DLB-SUM (second feature amount)) indicating the above is calculated, and the ratio of the total values (AD-SUM / DLB-SUM) is calculated as CIS. do. Further, when the Z value calculation unit 20 calculates the Z value, the voxel values (representative values) in the AD and DLB interest regions stored in the disease-specific interest region data storage unit 17 are obtained, and the voxel values (representative values) in the AD and DLB interest regions are obtained. The Z value may be calculated, and the feature amount (first feature amount, second feature amount) may be calculated based on the value. In this case, the feature amount (first feature amount, second feature amount) is stored in the Z value data storage unit 21, and the index calculation unit 22 is based on the feature amount stored in the Z value data storage unit 21. , The ratio of features is calculated as CIS.

The calculation data storage unit 23 stores the total value (AD-SUM, DLB-SUM) and CIS, which are the calculation data calculated by the index calculation unit 22.

The threshold data storage unit 24 stores threshold data indicating a predetermined threshold for distinguishing whether the subject is an AD patient or a DLB patient. The threshold data may be acquired via, for example, the input device 2 or the input / output device 4.

For the threshold value, the created AD interest region and DLB interest region are used to calculate the CIS for the AD patient group and the DLB patient group, and the threshold value is changed to distinguish between the AD patient and the DLB patient. When the sensitivity and specificity are calculated based on the number and the number of true negatives, the number of AD patients and the number of DLB patients, and ROC analysis (Receiving Operating Characteristic analogy) is performed, the sum of the sensitivity and the specificity becomes the largest. The threshold value may be a predetermined threshold value.

The discrimination unit 25 compares the CIS stored in the calculated data storage unit 23 with the threshold value stored in the threshold data storage unit 24, discriminates whether the subject is an AD patient or a DLB patient, and outputs a discrimination result. Output to 3.

The disease discrimination support device 1 of the present embodiment having the above-described configuration performs a disease-specific region of interest creation process and a subject discrimination process according to the procedure described below.

FIG. 6 is a flowchart showing a procedure for creating a disease-specific region of interest.

The AD patient data and the DLB patient data are stored in the AD patient data storage unit 11A and the DLB patient data storage unit 11B, respectively (S11). The AD and DLB patient data may be the normalized patient data performed in step S13, and in this case, step S13 is omitted.

Next, the normalization processing unit 13 normalizes the voxel value of the stored AD patient data and the voxel value of the DLB patient data, and calculates the standard deviation of the voxel value for each pixel based on the normalized voxel value. (S13).

The healthy person data is stored in the healthy person data storage unit 12 (S15). When the data to be saved in steps S11 and S15 has already been prepared, the corresponding steps may be omitted. After that, the normalization processing unit 13 normalizes the stored healthy person data, and calculates the standard deviation of the voxel value for each pixel based on the normalized healthy person data (S16).

The change area creation unit 14 is stored in the boxel value of the normalized AD patient data and DLB patient data, the standard deviation of the AD patient and DLB patient, the number of samples of the AD patient and DLB patient, and the healthy person data storage unit 12. A statistical test was performed for each pixel (boxel) based on the boxel value of the healthy person data, the standard deviation of the healthy person, and the number of samples of the healthy person, and the p value <0.05 (or 0.01). In some cases, it is assumed that there is a statistically significant difference, and a boxel determined to have a significant difference is created as a change region in which the blood flow is significantly changed (decreased) (S17). The created region is a disease-specific region (AD region and DLB region) of each disease, and the region is stored in the disease-specific region data storage unit 15. The change region creation unit 14 uses the first significance level and the second significance level (0.05, 0.01) to create the DLB region, and the data of the two change regions (second region data). , 2-2 area data) may be created.

The region of interest creation unit 16 compares the AD region and the DLB region stored in the disease-specific region data storage unit 15, and obtains the remaining region excluding the region where the AD region and the DLB region overlap from the AD region. As the AD area of interest, the DLB area is created as it is as the DLB area of interest (S19). Further, the created AD interest region and DLB interest region are stored in the disease-specific interest region data storage unit 17. When two change regions are created using the first significance level and the second significance level (0.05, 0.01) in the change region creation unit 14 regarding the creation of the DLB region, one of the changes is made. The region may be used to create the AD region of interest and the other region of change may be the DLB region of interest.

Then, after step S19, the disease-specific region of interest creation process is completed.

Next, the subject discrimination process will be described.

FIG. 7 is a flowchart showing a procedure for discriminating subjects.

First, the subject data is stored in the subject data storage unit 18 (S21).

Next, the normalization processing unit 19 normalizes the subject data stored in the subject data storage unit 18 (S23).

The Z value calculation unit 20 uses the boxel values (mean value and standard deviation) of the subject data normalized by the normalization processing unit 19 and the healthy subject data normalized by the normalization processing unit 13 to calculate the Z value. Calculate (S25). The calculated Z value is stored in the Z value data storage unit 21.

The index calculation unit 22 is positive for each disease with respect to the Z value stored in the Z value data storage unit 21 for each of the AD interest region and the DLB interest region stored in the disease-specific interest region data storage unit 17. The total value of the Z values indicating the values (AD-SUM (first feature amount), DLB-SUM (second feature amount)) is calculated, and the ratio thereof is used as an index (CIS: AD-SUM / DLB-SUM). Calculate (S27). The calculated CIS is stored in the calculated data storage unit 23.

The discrimination unit 25 compares the CIS stored in the calculated data storage unit 23 with the threshold value stored in the threshold data storage unit 24, discriminates whether the subject is an AD patient or a DLB patient, and outputs a discrimination result. Output to 3.

Next, the result of evaluating the discriminative ability between AD and DLB by the CIS calculated using the AD interest region and the DLB interest region created by the method of the present embodiment will be described. ..

As the AD area of interest and the DLB area of interest, the areas shown below were used.
-AD interest region N1: AD interest region N2 created by removing the region overlapping with the DLB region created with p value <0.01 from the AD region created with p-value (significance probability) <0.01 (significance level). : AD area created by excluding the area overlapping the DLB area created with p-value <0.05 from the AD area created with p-value <0.01 / DLB area of interest D1: Created with p-value <0.01 DLB region of interest D2: DLB region of interest created with a p-value <0.05

Thirteen AD patients and 13 DLB patients were differentiated. For the discrimination target, the Z value is calculated using the above AD and DLB areas of interest, the CIS is calculated, the threshold value is changed to distinguish between AD patients and DLB patients, and the number of true positives and true negatives at each threshold value. Sensitivity and specificity were calculated based on the number, the number of AD patients and the number of DLB patients, ROC analysis was performed, and evaluation was performed by AUC (Area under the curve). The evaluation results are shown in Table 1 below.
(Table 1)

As shown in Table 1, the AUC showed a value of 0.852 to 0.882, and a high discriminating ability was obtained as a whole. The highest discrimination ability was obtained with the combination of N2 / D1.

As described above, a region showing a significant decrease in blood flow in AD and a region in which no significant decrease in blood flow is observed in DLB is defined as an AD area of interest (disease-specific area of interest), and a significant decrease in blood flow in DLB. The Z value is calculated based on the subject data and the healthy subject data, and the ratio of the total of the positive Z values in the AD interest region and the DLB interest region is defined as CIS. By doing so, it is possible to provide a CIS showing a high discrimination ability.

The above-described embodiments of the present invention are examples for the purpose of explaining the present invention, and the scope of the present invention is not limited to those embodiments. One of ordinary skill in the art can practice the present invention in various other aspects without departing from the gist of the present invention.

For example, in the present embodiment, a tomographic image by SPECT is used, but a tomographic image by PET (Positron Emission Tomography), X-ray CT, or MRI may be used in addition to SPECT. The disease may be a disease other than AD and DLB.

Further, in the above embodiment, a region showing a significant decrease in blood flow in AD and a region in which no significant decrease in blood flow is observed in DLB is defined as an AD interest region (disease-specific interest region), and is significant in DLB. The Z value is calculated based on the subject data and the data of healthy subjects, and the total of the positive Z values in the AD interest region and the DLB interest region is calculated by setting the region where the decrease in blood flow is observed as the DLB interest region (disease-specific interest region). The ratio was CIS. However, the region in which a significant decrease in blood flow is observed in AD is defined as an AD area of interest (disease-specific area of interest), and the area in which DLB shows a significant decrease in blood flow, and a significant decrease in blood flow is observed in AD. The non-region may be the DLB region of interest (disease-specific region of interest), the Z value may be calculated based on the subject data and the healthy subject data, and the ratio of the total of the positive Z values in the AD interest region and the DLB interest region may be defined as CIS. .. That is, the remaining region excluding the region where the DLB region and the AD region overlap from the DLB region may be used as the DLB interest region, and the AD region may be used as it is as the AD interest region.

Further, in the above embodiment, the disease is differentiated based on the change in cerebral blood flow due to SPECT, but the disease may be differentiated based on the change in sugar metabolism due to PET. In this case, the change region creating unit 14 creates an AD region or the like based on the change in metabolism, creates an AD interest region or the like based on the region, and distinguishes the disease.

Further, in the above embodiment, the area of interest creation unit 16 creates the DLB area as it is as the DLB area of interest, but the remaining area excluding the area where the AD area and the DLB area overlap from the DLB area is the DLB area of interest. (Regions I and II in FIG. 4B) may be created. Further, although the discrimination unit 25 outputs the discrimination result to the output device, the discrimination unit 25 may output the CIS as an index in addition to or in place of the discrimination result, or may output the CIS and the threshold value.

Further, in the above embodiment, the change region creation unit 14 creates disease-specific region data indicating the disease-specific region (AD region and DLB region) of each disease based on all the regions where blood flow is decreased. The disease-specific region data storage unit 15 stores the disease-specific region data, and the region of interest creation unit 16 compares the AD region and the DLB region, and the AD region, the AD region, and the DLB region overlap each other. The remaining area excluding the area was created as an AD area of interest, but the present invention is not limited to this. For example, the change region creation unit 14 creates a disease-specific region (AD region and DLB region) of each disease, and then obtains AD patient data, DLB patient data, and healthy subject data for all pixels in each region. A statistical value may be calculated for each AD patient and DLB patient by comparison between the groups, and the statistical value and the coordinates of all the pixels in each region may be stored in the disease-specific region data storage unit 15. Then, the area of interest creation unit 16 may compare the statistical values of the pixels in the DLB area having the same coordinates among the pixels in the AD area, and create the AD area of interest based on the comparison result. .. For example, if the statistical value is a Z value, the Z value of an AD patient and the Z value of a DLB patient in pixels with the same coordinates are different, and if the value is smaller than a predetermined value, the data at that coordinate is used as the AD interest region. It may be excluded from.

1 Disease discrimination support device, 2 Input device, 3 Display device, 4 Input / output device, 10 Disease discrimination support device main unit, 11 Patient data storage unit, 12 Healthy person data storage unit, 13 Normalization processing unit, 14 Change area creation unit , 15 Disease-specific area data storage unit, 16 Interest area creation unit, 17 Disease-specific area of interest data storage unit, 18 Subject data storage unit, 19 Normalization processing unit, 20 Z value calculation unit, 21 Z value data storage unit , 22 Index calculation unit, 23 Calculation data storage unit, 24 Threshold data storage unit, 25 Identification unit

Claims (11)

Healthy person data which is data about image of a predetermined site of a large number of healthy persons, first patient data which is data about an image of a predetermined site of a large number of patients having a first disease, and a large number having a second disease. This is a computer program for causing a disease discrimination support device having a storage unit to store the second patient data, which is data related to the image of the predetermined portion of the patient, to be executed.
The healthy person data is compared with the first patient data to create a first disease data relating to the first disease at the predetermined site, and the healthy person data is compared with the second patient data. Then, the step of creating the second disease data regarding the second disease at the predetermined site, and
A step of correcting the first disease data based on the second disease data to create first region data indicating a disease-specific region of interest relating to the first specific disease .
A computer program for causing the disease discrimination support device to execute a step of creating second region data indicating a disease-specific region of interest related to the second specific disease based on the second disease data. The first disease data has data indicating a disease-specific region relating to the first disease, and the second disease data has data indicating a disease-specific region relating to the second disease.
In the step of creating the first area data,
The remaining region excluding the region where the disease-specific region of the first disease data and the disease-specific region of the second disease data overlap from the disease-specific region of the first disease data is the first region data. The computer program according to claim 1, which is created as. The computer according to claim 1 or 2, wherein the disease-specific region of interest of the first region data and the disease-specific region of interest of the second region data are partially in contact with each other without overlapping each other. program. In the step of creating the second disease data,
The disease group consisting of the second patient data and the healthy subject group consisting of the healthy subject data are compared between the groups, and a statistical test is performed for each pixel to show a disease-specific region based on the first significance level. The 2-1 disease data and the 2-2 disease data showing the disease-specific region based on the second significance level lower than the first significance level were created as the second disease data.
In the step of creating the first region data and the second region data,
The remaining region excluding the region where the disease-specific region of the first disease data and the disease-specific region of the 2-1 disease data overlap from the disease-specific region of the first disease data is the first. The computer program according to claim 2, wherein the computer program is created as area data, and the disease-specific area of the 2-2 disease data is used as the second area data. In the step of creating the second area data,
The remaining region excluding the region where the disease-specific region of the first disease data and the disease-specific region of the second disease data overlap from the disease-specific region of the second disease data is the second region data. The computer program according to claim 2, which is created as. According to the computer program according to any one of claims 1 to 5, the subject data which is the data related to the image of the predetermined part of the subject, the healthy person data which is the data related to the image of the predetermined part of a large number of healthy persons, and the computer program according to any one of claims 1 to 5. It is a computer program to be executed by a disease discrimination support device having a storage unit for storing the created first area data and second area data.
A step of comparing the subject data and the healthy subject data to calculate the feature amount,
The first feature amount calculated based on the feature amount of the disease-specific interest region of the first region data and the second feature amount calculated based on the feature amount of the disease-specific interest region of the second region data. Steps to calculate indicators based on
A computer program for causing the disease discrimination support device to execute a step of outputting the index. The computer program according to claim 6, wherein the feature amount is a Z value calculated based on the subject data and the healthy person data. The computer program according to any one of claims 1 to 7, wherein the image of the predetermined site is a SPECT (Single Photon Emission Tomography) tomography image of the brain or a PET (Positron Emission Tomography) tomography image. .. The computer program according to any one of claims 1 to 8, wherein the first disease is Alzheimer's disease and the second disease is Lewy body dementias. Healthy person data which is data about image of a predetermined site of a large number of healthy persons, first patient data which is data about an image of a predetermined site of a large number of patients having a first disease, and a large number having a second disease. A storage unit that stores the second patient data, which is data related to the image of the predetermined portion of the patient,
The healthy person data is compared with the first patient data to create a first disease data relating to the first disease at the predetermined site, and the healthy person data is compared with the second patient data. Then, a disease-specific data creation means for creating a second disease data relating to the second disease at the predetermined site, and
The first disease data is corrected based on the second disease data to create first region data indicating a disease-specific region of interest related to the first specific disease, and the first region data is created based on the second disease data. A disease discrimination support device comprising a disease-specific area of interest creating means for creating a second area data indicating a disease-specific area of interest relating to two specific diseases. Subject data which is data related to an image of a predetermined part of a subject, healthy person data which is data related to an image of a predetermined part of a large number of healthy persons, and first region data created by the disease discrimination support device according to claim 10. And a storage unit that stores the second area data,
A feature amount calculation means for calculating a feature amount by comparing the subject data with the healthy person data, and
The first feature amount calculated based on the feature amount of the disease-specific interest region of the first region data and the second feature amount calculated based on the feature amount of the disease-specific interest region of the second region data. Index calculation means for calculating indicators based on
A disease discrimination support device including an output means for outputting the index.

Images