JP2007280229A - Similar case retrieval device, similar case retrieval method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To retrieve a similar case considering a prevalence in an area where an image of a diagnostic object is obtained. <P>SOLUTION: Based on the incidence of each disease in a plurality of diseases incident in a specific area acquired from case data 100 stored in a case storage means 41 and the incidence of each disease corresponding to a value of characteristic quantity obtained from a case image 110 acquired in the specific area, a learning unit 42 learns the probability of each disease in the specific area of the case image 110. A prevalence acquisition means 43 acquires the probability of a diagnostic object image 130 being each disease by use of the learning unit 42 for the specific area where the diagnosis object image 130 is acquired. Further, case data 100 for the case image 110 similar to the diagnosis object image 130 is retrieved from the case data 100 stored in the case storage means 41 as similar case data 140. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a similar case search apparatus, a similar case search method, and a program for searching for cases similar to cases in medical images.

  In recent years, with the expansion of computer utilization technology, the introduction of electronic medical records using computers instead of paper medical records has been promoted for the purpose of improving the efficiency and speed of work in medical institutions. Furthermore, a system has been proposed in which electronic medical records are stored in a shared server that can be accessed by a plurality of medical institutions, and a plurality of medical institutions can share and use medical data of electronic medical records.

  However, the functions of the above-described system are merely obtained by storing electronic medical records from a plurality of medical institutions or providing the electronic medical records to a plurality of medical institutions, and the electronic medical records are not effectively used. It was. Therefore, it is proposed to calculate the rate of injury and illness in each region based on the location information indicating the location of the medical institution or patient from the electronic medical record, and to identify the region where many injuries and illnesses occur. (For example, patent document 1).

On the other hand, similar cases of diagnosis cases can be selected from various images such as distances between vectors such as decision trees, neural networks, linear / nonlinear classifiers, Mahalanobis distances, etc. There has been proposed a system that assists doctors' interpretation by searching for similar images with similar image characteristics using a technique and enabling reference to the similar images. Specifically, a method for searching similar cases for diffuse parenchymal lung disease (for example, Non-Patent Document 1) or a method for searching for similar cases for isolated shadows is proposed. (For example, Non-Patent Document 2).
JP-A-2004-287614 Aisen, AM, Broderick, LS, Winer-Muram, H., Brodley, CE, Kak, AC, and Pavlopoulou C., Dy, J., Marchiori, A., Automated storage and retrieval of medical images to assist diagnosis: Implementation and preliminary assessment, Radiology. 228 (1), pp. 265-270, July 2003 Li et al .: Investigation of new psychophysical measures., Medical Physics, Vol. 30, No. 10, October 2003

  However, in the case of a disease whose morbidity is biased depending on the region or hospital, if similar cases are searched using the same method, the search results may not match the actual situation, and results useful for diagnosis may not be obtained.

  For example, if there is a disease N (asbestosis, etc.) that occurs at a rate of 1 in 100 people in area A and 1 in 1000 in area B, a search created using case data in area A If the engine is used in the B area, the disease N having a low appearance frequency is searched for in the B area, and the search result may not match the actual situation.

  Therefore, in order to solve this problem, the present invention aims to propose a similar case search device, a similar case search method and a program thereof for searching for similar cases in consideration of the prevalence of each region. is there.

The similar case search device of the present invention includes a case storage means for storing case data including a case image acquired in a predetermined area and disease information obtained by diagnosing the case image;
From the case image of the case data and the disease information, the incidence of each disease among a plurality of diseases that occurred in the predetermined region, and the value of the feature value obtained from the case image acquired in the predetermined region A learning device that learns the probability that each case image is the disease in the one area based on the incidence of each disease corresponding to
Using the learning device, according to the feature amount acquired from the diagnosis target image, a disease rate acquisition means for obtaining a probability that the diagnosis target image is each disease of the plurality of diseases;
Similar case search means for searching, as the similar case data, case data of a case image close to the probability that the diagnosis target image is the disease from the case data stored in the case storage means It is characterized by.

Further, the similar case search method of the present invention includes a case storage step of storing case data including a case image acquired in a predetermined area and disease information obtained by diagnosing the case image,
From the case image of the case data and the disease information, the incidence of each disease among a plurality of diseases that occurred in the predetermined region, and the value of the feature value obtained from the case image acquired in the predetermined region A learning device generating step for generating a learning device that learns the probability that each case image is the disease in the one region based on the incidence of each disease corresponding to
Using the learning device, according to the feature amount acquired from the diagnosis target image, a disease rate acquisition step of obtaining a probability that the diagnosis target image is each disease in the plurality of diseases,
A similar case retrieval step of retrieving, as the similar case data, case data of a case image close to the probability that the diagnosis target image is the disease from the case data stored in the case storage means It is characterized by.

The program of the present invention is a computer,
Case storage means for storing case data including a case image acquired in a predetermined area and disease information obtained by diagnosing the case image;
From the case image of the case data and the disease information, the incidence of each disease among a plurality of diseases that occurred in the predetermined region, and the value of the feature value obtained from the case image acquired in the predetermined region A learning device that learns the probability that each case image is the disease in the one area based on the incidence of each disease corresponding to
Using the learning device, according to the feature amount acquired from the diagnosis target image, a disease rate acquisition means for obtaining a probability that the diagnosis target image is each disease in the plurality of diseases;
Functioning as similar case retrieval means for retrieving case data of case images close to the probability that the diagnosis target image is each disease from the case data stored in the case storage means as the similar case data. It is a feature.

  “Region” means a place in a delimited range, but there may be an overlapping place or a place not including a part. Moreover, what collected several areas may be used.

  “Disease information” is information related to a disease in a case image such as a disease name.

Further, the case storage means stores case data acquired in a plurality of regions separately for the region from which the case data was acquired,
The learning device learns using case data acquired in one of the plurality of regions among the case data stored in the case storage means, and the learning device is stored in each region. According to the plurality of similar case search device,
The disease rate acquisition means obtains the probability of each disease using a learning device according to the area from which the diagnostic target image is acquired from the plurality of learning devices,
The similar case retrieval unit may retrieve similar case data from the case data in the region where the diagnosis target image is acquired among the case data stored in the case storage unit.

  According to the present invention, diagnosis is performed using a learning device that learns the probability that a case acquired in a certain region is based on the incidence of each disease among a plurality of diseases that have occurred in a predetermined region. By obtaining the probability that the target image is each disease and searching for cases with case images close to the probability that the diagnosis target image is each disease as a similar case, an interpreter, such as a doctor, has a similar likelihood of each disease. Accurate interpretation can be performed by referring to the case image as a similar case.

  By generating multiple learners for each region, it is possible to search for disease probabilities and similar cases corresponding to multiple regions. By comparing these results, a more accurate diagnosis can be performed.

  Embodiments of a similar case retrieval apparatus according to the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of a medical system provided with a similar case retrieval apparatus according to an embodiment of the present invention. As shown in FIG. 1, a medical system 1 according to the present embodiment includes a modality 2 that captures a medical image, an image server 3 that stores a medical image obtained by capturing an affected area of a subject, and a workstation 4 on which a doctor interprets an image. The report server 5 stores a report created by the interpretation doctor interpreting the medical image displayed on the screen of the workstation 4, the electronic medical record server 6 that stores the electronic medical record, and the network 7.

  Modality 2 includes patients such as CR (Computed Radiography), CT (Computed Tomography), MRI (Magnetic Resonance Imaging), US (Ultrasound imaging), etc. It is a device or system for recording the image of the image as digital data.

  The image server 3 receives a large amount of medical images taken by each modality 2 installed in the radiology department via the network 7 and receives the medical images together with patient information on the subject, photographing date, information on the taken modality, and the like. Save in DICOM compliant format. In addition, the image server 3 is installed with database management software and has a function of searching for stored medical images using various information attached to the medical images.

  The workstation 4 has a function of reading out a medical image to be interpreted from medical images stored in the image server 3 via the network 7 and displaying it on the screen of the image display device. In order for an interpreter such as a doctor to interpret a medical image, the workstation 4 is preferably provided with a display device such as a high-definition CRT.

  The report server 5 stores the result of the medical image read by the image interpretation doctor with the image interpretation device 4 as an image interpretation report.

  The electronic medical record server 6 stores the electronic medical record of each patient. In the electronic medical record, medical date and time, input person, medical record, various findings, test results, instruction record, insurance score claim items, and the like are recorded. Furthermore, order information of various examinations such as image photographing is also recorded, and medical images obtained by photographing each patient stored in the image server 3 and pathological result information such as biological examinations can be obtained via the ordering system.

  In addition, a similar case search program is installed in the workstation 4 so as to function as the similar case search device 40. The case data 100 similar to the diagnosis target image is displayed on the screen of the workstation 4 together with the diagnosis target image. Is displayed.

  As illustrated in FIG. 2, the similar case search device 40 includes a case storage unit 41 that stores a plurality of case data 100 divided into acquired regions, a case image 110 of the case data 100 acquired in each region, and a disease From the information 120, the incidence of each disease among a plurality of diseases that occurred in the region, and each disease corresponding to the feature value obtained from the case image 110 acquired in the specific region for each disease The learning device 42 that learns the occurrence probability and outputs the probability that the case image 110 is a disease in each region according to each region, and the region where the diagnosis target image 130 is acquired from the plurality of learning devices 42 Of the case images 110 of the case data 100 acquired in a specific region, the disease rate acquisition means 43 for obtaining the probability of being each disease among the diseases having a plurality of diagnosis target images 130 using the corresponding learning device 42 , Diagnostic vs Includes a similar case retrieval means 44 for retrieving case data 100 of the case image 110 which is similar to the image 130 as the similar case data 140, a display unit 45 for displaying the probability and the similar case data 140 is each disease.

  Case data 100 includes a case image 110 and disease information 120 of the case image 110. The case image 110 of the case data 100 is selected from medical images stored in the image server 3. The disease information 120 includes information on the confirmed disease name. The disease name is based on the interpretation report of the case image 110 stored in the report server 5, the medical record of the electronic medical record corresponding to the case image 110 stored in the electronic medical record server 6 (including definitive diagnosis and various findings), and the like. can get. Further, the disease information 120 may include information (for example, link information) related to the storage location where the electronic medical record or the interpretation report is stored.

  In selecting the case image 110, based on the address of the hospital where the case image 110 was taken and the address where the patient lives, an interpretation report of the case image 110 and an electronic medical record corresponding to the case image 110 are provided for each region. The medical record, various findings, and the like are referred to, and what is considered to be effective for comparison with the diagnosis target image 130 at the time of diagnosis is selected and stored in the case storage unit 41. The case data 100 is preferably stored in separate storage units (folders, partitions, storage devices, etc.) for each region.

  The learning device 42 corresponds to the incidence rate of each disease among a plurality of diseases occurring in a specific region and the feature value obtained from the case image 110 in which each disease is acquired in the specific region. Based on the occurrence probability of each disease, the probability that the case image 110 is each disease is learned.

In the conventional image search, the similarity between the diagnosis target image and each case image in the case database is calculated using the feature amount of the image, and searches are performed as similar cases in descending order of similarity. If the disease has a bias in incidence (for example, asbestos lung, farmer's lung, etc.), even if a similar case is searched using a learning device that has learned the same case data, the search results do not match the actual condition, Useful results may not be obtained.
Therefore, using the case data 100 of a specific area stored in the case storage means 41, the incidence rate of each disease occurring in that area is calculated. Further, a feature amount suitable for discriminating each disease is calculated from the case image 110 in the region, and each disease occurring in the region is calculated based on the value of the feature amount and the disease obtained from the disease information 120. Is calculated according to the feature value. For example, in the case where a feature value representing ground glass shading and nodular shadow is used, the larger the value of the feature amount representing ground glass shading, the smaller the incidence of lung cancer and the higher the incidence of asbestosis. On the other hand, the greater the value of the feature amount representing the nodule shadow, the smaller the incidence of asbestosis and the higher the incidence of lung cancer. As described above, which disease is suitable for discrimination depending on each feature amount, it is desirable to calculate the incidence of each disease according to the value of each feature amount.

  The learning device 42 learns the incidence rate of each disease occurring in a specific area and the incidence rate of each disease according to the feature value of the case image 110 acquired in the specific area. The probability that the image 110 is each disease can be output.

  Specifically, a method using Bayes' theorem is shown below as an example of creating a learning device that takes the occurrence rate into consideration. Here, a case where a similar case search for three diseases of lung cancer, interstitial pneumonia, and asbestosis is performed will be described as an example. For example, in the A region and the B region, three diseases occur at a frequency as shown in FIG. 3, and nodule shadows and ground glass shadows are used as feature amounts of the case image 110 for calculating the similarity of each disease. For the sake of simplicity, here, the feature amount corresponding to each of the nodule shadow and the ground glass shadow is assumed to be one, and the feature amount value is also assumed to be binary (1 when corresponding to the shadow, and 0 when not corresponding). Actually, the feature quantity used for discrimination is not limited to two, and there are a plurality of feature quantities, and the feature quantity values are also multivalued. The number of cases that occurred in each region is as shown in FIG. 3, and region A has a higher probability of asbestosis than region B.

Assume that the disease name is ωi, the feature quantity obtained from the case image 110 is xj, and the probability that the case image 110 having a certain feature quantity is a predetermined disease is P (ωi | xj). Also, assuming that the probability of each disease according to the value of a specific feature value is P (xj | ωi), the probability that the disease is ωi when the feature value obtained from the case image 110 is xj based on the Bayes' theorem. Can be calculated as follows.

  Therefore, the incidence rate of each disease in a specific area and the incidence rate of each disease according to the feature value of the case image 110 of the case data 100 obtained in that area are based on the Bayes' theorem of equation (1). Each learner 42 is trained to satisfy. In addition, for each region where the case data 100 is acquired, the incidence of each disease in each region, and the incidence of each disease according to the feature value of the case image 110 of the case data 100 obtained in that region The learning device 42 is generated corresponding to a plurality of regions for each region.

  The disease rate acquisition unit 43 acquires a feature amount from the diagnosis target image 130 to be diagnosed, and uses the learning device 42 in the area where the diagnosis target image 130 is acquired, and sends it to the learning device 42 from the diagnosis target image 130. The acquired feature amount is input to obtain a probability that the diagnosis target image 130 is each disease.

  When there is a relationship between the number of occurrences of each disease and the feature amount as shown in FIG. 3, the learning device 42 created for the A region, the learning device 42 created for the B region, and the learning device created for the A + B region In 42, the following results are obtained.

1) In the learning device 42 created using the incidence rate in the A area, when the diagnosis target image 130 has the characteristics of the ground glass shadow, the respective probabilities P (diseases) are cancer, interstitial pneumonia, and asbestosis. Lung cancer | Ground glass, P (Interstitial pneumonia | Ground glass), P (Asbestos lung | Ground glass) and the diagnosis target image 130 have the characteristics of ground nodule, the disease is cancer, interstitial pneumonia, asbestos Each probability P (lung cancer | nodule shadow), P (interstitial pneumonia | nodule shadow), and P (asbestosis | nodule shadow) of the lung has the following results.

2) Similarly, in the learning device 42 created using the incidence of the B region, when the diagnosis target image 130 has the characteristics of the ground glass shadow, the disease is cancer, interstitial pneumonia, asbestosis. Probability P (lung cancer | ground glass), P (interstitial pneumonia | ground glass), P (asbestos lung | ground glass) and the diagnosis target image 130 have the characteristics of ground nodule shadow, the disease is cancer, interstitial Each probability P (lung cancer | nodule shadow), P (interstitial pneumonia | nodule shadow), P (asbestosis | nodule shadow) of pneumonia and asbestosis has the following results.

3) Furthermore, in the learning device 42 created using the incidence of the A + B area, when the diagnosis target image 130 has the characteristics of the ground glass shadow, the disease is cancer, interstitial pneumonia, asbestosis Probability P (lung cancer | ground glass), P (interstitial pneumonia | ground glass), P (asbestosis | ground glass), and the diagnosis target image 130 has the characteristics of ground nodule shadow, the disease is cancer, stroma Probability P (lung cancer | nodule shadow), P (interstitial pneumonia | nodule shadow), and P (asbestosis | nodule shadow) for Pneumonia and asbestosis are as follows.

  The similar case retrieval unit 44 resembles the case data 100 of the case image 110 similar to the diagnosis target image 130 from the case images 110 of the case data 100 acquired in a specific area stored in the case storage unit 41. Search as case data 140.

  Whether or not the diagnosis target image 130 and the case image 110 are similar is obtained by determining that the incidence of each disease is similar. For example, in region A, the probability of each disease in the diagnosis target image 130 having the characteristics of ground glass is 0.22 for lung cancer, 0.38 for interstitial pneumonia, and 0.40 for asbestosis. Therefore, from the case data 100 acquired in the A area, the case data 100 of the case image 110 having a similar incidence of each disease in the diagnosis target image 130 is searched as the similar case data 140. Alternatively, the diagnosis target image 130 and the case image 110 having similar feature quantities may be similar.

  Similarly, when the probability of each disease in the diagnosis target image 130 having the characteristics of ground glass is obtained in the B area, the case image having a high similarity to the diagnosis target image 130 from the case data 100 acquired in the B area. A plurality of case data 100 of 110 are detected as similar case data 140.

  The display means 45 displays the probability of each disease in the diagnosis target image obtained by the disease rate obtaining means 43, the case image 110 of the similar case data 140 retrieved by the similar case retrieval means 44, and the disease obtained from the disease information 120. Displays the name. If necessary, the electronic medical record medical record or interpretation report of the case may be displayed from a storage location such as the electronic medical record or interpretation report included in the disease information 120.

  Next, a flow of searching for a similar image similar to the diagnosis target image and interpreting the diagnosis target image will be described based on the flowchart of FIG.

  First, the learning device 42 provided in the workstation 4 generates the learning device 42 for each region using the case data 100 of each region stored in the case storage unit 41. Each learning device 42 includes the occurrence rate of each disease corresponding to the incidence of each disease among a plurality of diseases occurring in each region and the feature value obtained from the case image 110 acquired in each region. Based on the probability, the probability that the case image 110 is each disease is learned. (S100). At this time, the learning device 42 for the whole country is generated using all the case data 100 stored in the case storage means 41.

  When the interpretation doctor performs interpretation, first, a list of medical images to be diagnosed is displayed on the screen of the workstation 4. When the interpretation doctor designates the diagnosis target image 130 from the list, the image server 3 is displayed. The diagnosis target image 130 is retrieved from the list and displayed on the display device (S101). The image interpretation doctor observes the displayed diagnosis object image 130 and creates an image interpretation report. When the image interpretation is performed, the disease rate acquisition unit 43 obtains a feature amount from the diagnosis object image 130, and the diagnosis object image is obtained. Using the local learning device 42 that has acquired 130, the feature amount acquired from the diagnosis target image 130 is input to the learning device 42, the probability that the diagnosis target image 130 is each disease is obtained, and the display means 45 displays the screen. The probability of each disease is displayed together with the diagnosis target image 130 (see FIG. 5).

  Further, the similar case retrieval unit 44 retrieves the case data 100 similar to the diagnosis target image 130. When the search for the similar case data 140 is instructed, the workstation 4 searches for the case image 110 similar to the diagnosis target image 130 from the case data 100 of the area where the diagnosis target image 130 is acquired (S102). Then, the display unit 45 displays the case data 100 of the case image 110 having a high similarity as the similar case data 140 as shown in FIG. 5 (S103).

  FIG. 5A shows an example in which a case image 110 of case data 100 similar to the diagnosis target image 130 having the characteristics of ground glass acquired in the A area is displayed on the screen. The incidence rate of the disease in the case image 110 of the similar case data 100 detected by the similar case search unit 44 is obtained using the learning device 42 for the A region, and the similar case data 140 is searched in descending order of the probability of each disease. The case image 110 is displayed together with the disease name. The diagnosis target image 130 having the characteristics of ground glass acquired in the area A is displayed in the order of “asbestosis”, “interstitial pneumonia”, and “lung cancer”.

  However, even if the diagnosis target image 130 has similar characteristics, when acquired in the region B, as shown in FIG. 5 (b), “interstitial pneumonia”, “lung cancer”, “asbestos” It is displayed in the order of “lung”. Thereby, since the case data 100 corresponding to the disease is displayed in the order corresponding to each region, it becomes easy to compare with each disease.

  In addition, the probability that the diagnosis target image 130 is each disease is obtained by using the regional learning device 42 from which the diagnosis target image 130 is acquired, and the national learning device 42 is used to determine the probability of each disease. In order to display the probability of each disease in and compare with the national probability, it may be displayed as shown in FIG.

  Alternatively, only when the disease search order is different between the whole country and the region, both may be displayed together to alert the doctor.

  The interpretation doctor compares the case image 110 of the similar case data 140 displayed on the screen with the diagnosis target image 130, confirms the disease name of the similar case data 140, and further records the medical record of the electronic medical record of the similar case And display means 45 such as an interpretation report may be displayed for confirmation. The interpretation doctor performs interpretation with reference to the similar case data 140, and inputs the interpretation result as an interpretation report (S104). When the input of the interpretation report is completed, the interpretation report is transmitted to the report server 5 and stored (S105).

  In the above description, the case data is collected for each region and a plurality of learning devices corresponding to each region are prepared. However, only one learning device is prepared by collecting case data only for a specific region. Thus, diagnosis can be performed according to the region with a learning device according to the region. Specifically, by creating a learning device using case data acquired at one hospital, a diagnosis suitable for the location of the hospital can be performed.

    As described above in detail, by using a learning device that learns the incidence of disease using the case data of the area where the image was acquired, the case corresponding to the actual situation of the area where the image to be diagnosed was acquired Data can be referred to, and diagnostic accuracy can be improved.

Block diagram showing schematic configuration of medical system Block diagram showing schematic configuration of similar case search device Frequency diagram showing the frequency of occurrence of each disease by region Flow chart showing the flow of searching for similar case images similar to the diagnosis target image and interpreting the diagnosis target image An example of how to display similar cases

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Medical system 2 Modality 3 Image server 4 Workstation 5 Report server 6 Electronic medical record server 7 Network 40 Similar case search device 41 Case memory means 42 Learning device 43 Disease rate acquisition means 44 Similar case search means
100 case data
110 case images
120 Disease information
130 Images to be diagnosed
140 Similar case data

Claims (4)

Case storage means for storing case data including a case image acquired in a predetermined area and disease information obtained by diagnosing the case image;
From the case image of the case data and the disease information, the incidence of each disease among a plurality of diseases that occurred in the predetermined region, and the value of the feature value obtained from the case image acquired in the predetermined region A learning device that learns the probability that each case image is the disease in the one area based on the incidence of each disease corresponding to
Using the learning device, according to the feature amount acquired from the diagnosis target image, a disease rate acquisition means for obtaining a probability that the diagnosis target image is each disease of the plurality of diseases;
Similar case search means for searching, as the similar case data, case data of a case image close to the probability that the diagnosis target image is the disease from the case data stored in the case storage means A similar case retrieval apparatus characterized by The case storage means stores case data acquired in a plurality of regions separately for the region from which the case data was acquired,
The learning device learns using case data acquired in one of the plurality of regions among the case data stored in the case storage means, and the learning device is stored in each region. According to the plurality of similar case search device,
The disease rate acquisition means obtains the probability of each disease using a learning device according to the area from which the diagnostic target image is acquired from the plurality of learning devices,
2. The similar case according to claim 1, wherein the similar case retrieval unit retrieves similar case data from case data in a region where the diagnosis target image is acquired from among the case data stored in the case storage unit. Search device. A case storage step for storing case data including a case image acquired in a predetermined area and disease information obtained by diagnosing the case image;
From the case image of the case data and the disease information, the incidence of each disease among a plurality of diseases that occurred in the predetermined region, and the value of the feature value obtained from the case image acquired in the predetermined region A learning device generating step for generating a learning device that learns the probability that each case image is the disease in the one region based on the incidence of each disease corresponding to
Using the learning device, according to the feature amount acquired from the diagnosis target image, a disease rate acquisition step of obtaining a probability that the diagnosis target image is each disease in the plurality of diseases,
A similar case retrieval step of retrieving, as the similar case data, case data of a case image close to the probability that the diagnosis target image is the disease from the case data stored in the case storage means A similar case search method characterized by Computer
Case storage means for storing case data including a case image acquired in a predetermined area and disease information obtained by diagnosing the case image;
From the case image of the case data and the disease information, the incidence of each disease among a plurality of diseases that occurred in the predetermined region, and the value of the feature value obtained from the case image acquired in the predetermined region A learning device that learns the probability that each case image is the disease in the one area based on the incidence of each disease corresponding to
Using the learning device, according to the feature amount acquired from the diagnosis target image, a disease rate acquisition means for obtaining a probability that the diagnosis target image is each disease of the plurality of diseases;
Functioning as similar case retrieval means for retrieving case data of case images close to the probability that the diagnosis target image is the disease from the case data stored in the case storage means as the similar case data. A featured program.

Images