Medical image diagnostic system, method for operating medical image diagnostic system, and program

Abstract

To provide a medical image diagnostic system, a medical image diagnostic method and a program for reducing the labor of a doctor in the case of performing image diagnosis to a large amount of medical images such as medical checkup.SOLUTION: At least one processor and at least one memory storing instructions for execution by the at least one processor, wherein the at least one processor is configured to perform first determination for determining whether or not a medical image obtained by imaging a subject is normal, and perform second determination for determining whether or not there is an abnormality from the medical image in a case where it is determined that the medical image is not normal in the first determination.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a medical image diagnostic system, a medical image diagnostic method, and a program, and more particularly to a technique for supporting the diagnosis of medical images. [Background technology]

[0002] Systems that use AI (Artificial Intelligence) to detect abnormal regions in medical images and assist in diagnosis are known.

[0003] For example, Patent Document 1 discloses a medical image processing system that, regarding abnormal shadow candidates detected from medical images, provides only detection information on abnormal shadow candidates that are questionable as to whether they are true positive abnormal shadows and / or abnormal shadow candidates with low visibility, thereby preventing doctors from overlooking abnormal shadows and improving the efficiency of image interpretation work. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2006-340835 Summary of the Invention [Problem to be solved by the invention]

[0005] For health management, medical checkups are conducted to examine the health status of subjects. In medical checkups, the subjects are mainly healthy individuals, and the diagnostic purpose, target organs, and target diseases are limited. When a possible abnormality is found in a medical image obtained during a medical checkup, a doctor must determine whether it is truly abnormal, diagnose the possible disease, and prepare a detailed report, all of which are important tasks.

[0006] On the other hand, doctors must also review images that show no abnormalities. For example, in health checkups for young people, the likelihood of patients with abnormalities is low, so doctors must review a large number of "images that show no abnormalities," placing a heavy burden on doctors.

[0007] In response to this, it is conceivable to support image diagnosis with lesion detection AI. However, lesion detection AI is typically created for each disease, and with the vast number of disease types, it is difficult to create AI that can handle all diseases. Furthermore, the amount of training data for rare diseases is small, making it difficult to create lesion detection AI. Furthermore, it is not possible to create lesion detection AI for unknown diseases, but doctors are still responsible for diagnosing unknown diseases. As such, even if the number of lesion detection AIs increases and their accuracy improves, there is a limit to the accuracy of the final lesion detection AI output, which has been an issue.

[0008] Furthermore, considering the number of target diseases and manufacturers, the lineup of lesion detection AI is enormous, and therefore, running all the existing lesion detection AI processes on all input images is difficult in terms of hardware resources and processing time.

[0009] The present invention has been made in consideration of the above circumstances, and aims to provide a medical image diagnostic system, a medical image diagnostic method, and a program that reduce the burden on doctors when performing image diagnosis on a large number of medical images, such as in health checkups. [Means for solving the problem]

[0010] One aspect of a medical image diagnostic system for achieving the above object is a medical image diagnostic system comprising at least one processor and at least one memory for storing instructions to be executed by the at least one processor, wherein the at least one processor performs a first determination to determine whether a medical image obtained by photographing a subject is normal, and, if the medical image is determined to be abnormal in the first determination, performs a second determination to determine whether an abnormality is present in the medical image. A medical image is normal when, for example, the medical image can be considered an image of a healthy person. A healthy person is a healthy person, for example, one who does not have any disease, illness, or lesion. This aspect can reduce the burden on doctors when performing image diagnosis on a large number of medical images.

[0011] Preferably, when the medical image is determined to be normal in the first determination, the at least one processor performs a third determination of determining whether or not an abnormality is present in the medical image.

[0012] Preferably, the at least one processor performs the third determination at a timing different from that of the second determination.

[0013] It is preferable that at least one processor makes a first determination using the probability that the medical image is normal, and makes a third determination if the medical image is determined to be normal in the first determination with a probability lower than a first threshold.

[0014] It is preferable that at least one processor performs a first discrimination using a first trained model that, when a medical image is input, outputs whether the medical image is normal or not, and re-trains the first trained model using a medical image that is discriminated to be abnormal in the third discrimination.

[0015] The first trained model is preferably a trained model trained using a training dataset consisting of a combination of medical images with and without abnormalities and a label indicating the presence or absence of an abnormality.

[0016] If the second discrimination determines that no abnormality is present, at least one processor performs a fourth discrimination to determine whether or not an abnormality is present from the medical image, and it is preferable that in the fourth discrimination, the presence or absence of an abnormality is determined from the medical image with a sensitivity relatively higher than the sensitivity in the second discrimination.

[0017] It is preferable that at least one processor displays the medical image discrimination results on the display in different manners for the first case in which the medical image is discriminated to be normal in the first discrimination, the second case in which the medical image is discriminated to be abnormal in the first discrimination and is discriminated to have an abnormality in the second discrimination, and the third case in which the medical image is discriminated to be abnormal in the first discrimination and is discriminated to have no abnormality in the second discrimination.

[0018] Preferably, the at least one processor displays the discrimination results of the medical image on the display in different modes for the second and third cases.

[0019] Preferably, the at least one processor performs different post-processing on the medical image in the first, second and third cases.

[0020] Preferably, the at least one processor performs a first discrimination and a second discrimination for each organ of the subject from the medical image.

[0021] Preferably, the at least one processor performs the second discrimination using a second trained model that receives a medical image as input and outputs an abnormality in the medical image.

[0022] One aspect of a medical image diagnostic method for achieving the above object is a medical image diagnostic method comprising: a first determination step of determining whether a medical image obtained by photographing a subject is normal; and a second determination step of determining whether or not an abnormality exists in the medical image if the medical image is determined to be abnormal in the first determination step. According to this aspect, it is possible to reduce the burden on doctors when performing image diagnosis on a large number of medical images.

[0023] One aspect of a program for achieving the above object is a program for causing a computer to execute the above medical image diagnostic method. This aspect may also include a computer-readable non-transitory storage medium on which this program is recorded. [Effects of the Invention]

[0024] According to the present invention, it is possible to reduce the burden on doctors when performing image diagnosis on a large number of medical images. [Brief explanation of the drawings]

[0025] [Figure 1] FIG. 1 is a block diagram of a medical image diagnostic system according to this embodiment. [Figure 2] FIG. 2 is a flowchart showing a medical image diagnosis method. [Figure 3] FIG. 3 is a process diagram illustrating a medical image diagnosis method. [Figure 4] FIG. 4 is a diagram showing a display form. [Figure 5] FIG. 5 is a diagram showing a display form. [Figure 6] FIG. 6 is a diagram showing a display form. [Figure 7] FIG. 7 is a diagram showing a display form. DETAILED DESCRIPTION OF THE INVENTION

[0026] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0027] [Configuration of medical image diagnostic system (medical image diagnostic support system)] The medical image diagnostic system of this embodiment reduces the burden on doctors when determining whether or not there are abnormalities in a large number of medical images, such as in health checkups, and determining whether or not the medical images are normal (whether or not they are images of healthy individuals).

[0028] Fig. 1 is a block diagram of a medical image diagnostic system 10 according to this embodiment. As shown in Fig. 1, the medical image diagnostic system 10 includes a modality 12, an image storage server 14, a CAD (Computer-Aided Diagnosis) processing server 18, a result integration CAD processing server 16, and a PACS (Picture Archiving and Communication System) viewer 20.

[0029] The modality 12, image storage server 14, each company's CAD processing server 18, result integration CAD processing server 16, and PACS viewer 20 are each connected to a communication network such as the Internet so as to be able to send and receive data.

[0030] The modality 12 is an imaging device that captures an image of a target area of ​​a subject and generates a medical image. The modality 12 includes, for example, at least one of an X-ray imaging device, a CT (Computed Tomography) device, an MRI (Magnetic Resonance Imaging) device, a PET (Positron Emission Tomography) device, an ultrasound device, and a CR (Computed Radiography) device using a flat X-ray detector.

[0031] The image storage server 14 is a server that manages medical images captured by the modality 12. The image storage server 14 is implemented as a computer equipped with a large-capacity storage device. Software that provides the functions of a data storage system is installed in the computer. The image storage server 14 acquires medical images captured by the modality 12 and stores them in the large-capacity storage device.

[0032] The format of a medical image can be in accordance with the DICOM (Digital Imaging and Communications in Medicine) standard. DICOM tag information defined in the DICOM standard may be added to the medical image. Note that the term "image" in this specification can refer to not only the image itself, such as a photograph, but also image data, which is a signal representing an image.

[0033] The result integrated CAD processing server 16 includes a first discrimination unit 16A. The first discrimination unit 16A performs a first discrimination on medical images acquired from the image storage server 14, discriminating whether the medical image is normal for each organ. Whether a medical image is normal or not refers, for example, to whether the medical image can be said to be an image of a healthy person. A healthy person is a healthy person, and is, for example, one who does not have any disease, illness, or lesion. The first discrimination result by the first discrimination unit 16A is linked to the medical image in the image storage server 14 and stored in a large-capacity storage device.

[0034] The result integrated CAD processing server 16 is implemented as a personal computer or a workstation. The result integrated CAD processing server 16 includes a processor 16B and a memory 16C. The processor 16B executes instructions stored in the memory 16C.

[0035] The hardware structure of processor 16B is various processors as follows: The various processors include a CPU (Central Processing Unit), which is a general-purpose processor that executes software (programs) and functions as various functional units, a GPU (Graphics Processing Unit), which is a processor specialized for image processing, a PLD (Programmable Logic Device), which is a processor whose circuit configuration can be changed after manufacturing such as an FPGA (Field Programmable Gate Array), and a dedicated electrical circuit, which is a processor having a circuit configuration designed specifically for executing specific processing such as an ASIC (Application Specific Integrated Circuit).

[0036] A single processing unit may be configured with one of these various processors, or may be configured with two or more processors of the same or different types (e.g., multiple FPGAs, a combination of a CPU and an FPGA, or a combination of a CPU and a GPU). Also, multiple functional units may be configured with a single processor. Examples of multiple functional units configured with a single processor include, first, a configuration in which a single processor is configured with a combination of one or more CPUs and software, as typified by a client or server computer, and this processor operates as multiple functional units. Second, a configuration in which a processor is used to realize the functions of an entire system including multiple functional units on a single IC (Integrated Circuit) chip, as typified by an SoC (System On Chip). In this way, the various functional units are configured with one or more of the above-mentioned various processors as a hardware structure.

[0037] Furthermore, the hardware structure of these various processors is, more specifically, an electric circuit made up of a combination of circuit elements such as semiconductor elements.

[0038] The memory 16C stores instructions to be executed by the processor 16B. The memory 16C includes a RAM (Random Access Memory) and a ROM (Read Only Memory), not shown. The processor 16B uses the RAM as a working area and executes software using various programs and parameters, including a medical image processing program, stored in the ROM, and as a result, executes various processes of the integrated CAD processing server 16 by using the parameters stored in the ROM, etc.

[0039] The respective company CAD processing server 18 is composed of a plurality of CAD processing servers owned by a plurality of companies. The respective company CAD processing server 18 may be a single CAD processing server. The respective company CAD processing server 18 includes a second discrimination unit 18A. The second discrimination unit 18A includes a program for performing a second discrimination, which executes an abnormality detection process for each organ on medical images acquired from the image storage server 14 and which have been discriminated as abnormal at least by the first discrimination unit 16A, and discriminates the presence or absence of one or more abnormalities from the medical images. An abnormality includes, for example, at least one of a disease, a disorder, and a lesion. The second discrimination result of the second discrimination unit 18A is linked to the medical image in the image storage server 14 and stored in a large-capacity storage device.

[0040] The second discrimination unit 18A may be provided in the result integration CAD processing server 16.

[0041] The PACS viewer 20 is a terminal device used by a user such as a doctor, and may be, for example, a known image viewer for image interpretation. The PACS viewer 20 may be a personal computer, a workstation, or a tablet terminal.

[0042] The PACS viewer 20 includes an input device 20A and a display 20B. The input device 20A includes a pointing device such as a mouse and an input device such as a keyboard. A user can input instructions to the medical image diagnostic system 10 using the input device 20A. The display 20B displays a screen required for operations on the input device 20A and functions as a part that realizes a GUI (Graphical User Interface). The display 20B displays medical images captured by the modality 12. The display 20B also displays the first discrimination result and the second discrimination result as CAD results. The PACS viewer 20 may be a touch panel display in which the input device 20A and the display 20B are integrated.

[0043] [Medical image diagnosis method (medical image diagnosis support method)] FIG. 2 is a flowchart showing a medical image diagnosis method using the medical image diagnostic system 10. FIG. 3 is a process diagram showing the medical image diagnosis method. The medical image diagnosis method is realized by the processor 16B executing a medical image diagnosis program (medical image diagnosis support program) stored in the memory 16C. When the second discrimination unit 18A is provided in the respective company's CAD processing server 18, the medical image diagnosis method is realized by a processor (not shown) provided in the respective company's CAD processing server 18 executing the medical image diagnosis program stored in a memory (not shown) provided in the respective company's CAD processing server 18, and the processor 16B executing the medical image diagnosis program stored in the memory 16C. The medical image diagnosis program may be provided by a computer-readable non-transitory storage medium. In this case, the result integrated CAD processing server 16 may read the medical image diagnosis program from the non-transitory storage medium and store it in the memory 16C.

[0044] The medical image diagnosis method is performed for each organ of a subject. Here, the case of diagnosing a CT image of the lungs will be explained as an example.

[0045] In step S1, the processor 16B of the result integrated CAD processing server 16 causes the image storage server 14 to acquire CT images of the subject's lungs taken by the modality 12. The image storage server 14 acquires the CT images taken by the modality 12.

[0046] In step S2 (an example of a "first discrimination step"), the processor 16B inputs the CT image acquired by the image storage server 14 to the first discrimination unit 16A. As shown in Fig. 3, the first discrimination unit 16A includes a normality discrimination AI 16D that discriminates whether the medical image is normal or not.

[0047] The normality discrimination AI16D is a trained model (an example of a "first trained model") that, when a lung CT image is input, determines whether the CT image is normal or not, and includes a convolutional neural network. The normality discrimination AI16D is generated by deep learning using a training dataset of medical images and labels indicating the presence or absence of abnormalities. For example, the normality discrimination AI16D is generated by deep learning using a training dataset of normal CT images and normal labels, and a training dataset of abnormal CT images and abnormal labels. Note that a normal CT image is a CT image of a healthy subject. An abnormal CT image is a CT image with some kind of abnormality, such as a CT image of a subject with at least one of a disease, illness, or lesion.

[0048] The normality determination AI16D is generated to output the degree of normality of the input CT image as a numerical value (a score, an example of "probability"). The normality determination AI16D outputs that the CT image is not normal if the normality of the CT image is below a predetermined threshold, and outputs that the CT image is normal if the normality of the CT image is equal to or greater than the threshold. The first determination unit 16A inputs the CT image to the normality determination AI16D and performs a first determination (process P1).

[0049] In step S3, processor 16B acquires the first discrimination result from first discrimination unit 16A and determines whether the CT image is normal based on the first discrimination result. That is, processor 16B determines whether there is an abnormality in the CT image. If the CT image is normal (an example of the "first case", process P2), the process proceeds to step S4, and if the CT image is not normal (process P3), the process proceeds to step S5.

[0050] In step S4, the processor 16B displays the normal CT image on the display 20B of the PACS viewer 20 in display format A, and performs post-processing of processing format A (process P4). Furthermore, the processor 16B assigns "type A" incidental information to the CT image and stores it in the image storage server 14, and ends the processing of this flowchart. The post-processing of processing format A is an example of post-processing in the first case where the medical image is determined to be normal in the first determination.

[0051] In step S5 (an example of a "second discrimination step"), the processor 16B inputs the abnormal CT image to the company's CAD processing server 18. The company's CAD processing server 18 inputs the CT image to the second discrimination unit 18A and obtains the second discrimination result (process P5).

[0052] 3, the second discrimination unit 18A includes a company A lesion detection AI 18B that detects disease α, a company A lesion detection AI 18C that detects disease β, a company A lesion detection AI 18D that detects disease γ, a company B lesion detection AI 18E that detects disease γ, and a company C lesion detection AI 18F that detects disease β. For example, disease α is lung cancer, disease β is pneumonia, and disease γ is pneumothorax.

[0053] Each of the lesion detection AIs 18B-18F is a trained model (an example of a "second trained model") that outputs diseased areas (lesion areas, examples of "abnormalities") within a CT image of the lungs when the image is input, and each includes a convolutional neural network. Each of the lesion detection AIs 18B-18F is generated by deep learning using labeled images in which a doctor labels each diseased area within the CT image of the lungs as training data.

[0054] Each lesion detection AI 18B-18F is configured to calculate the probability of each disease for each pixel in the CT image, and pixels exceeding a predetermined threshold are considered to be diseased areas. Each lesion detection AI 18B-18F has higher specificity than when each AI performs lesion detection alone, i.e., the threshold is set relatively high. This allows each lesion detection AI 18B-18F to detect areas with a higher probability of being a lesion. This is because abnormalities are rare in health checkups, etc.

[0055] Second discrimination unit 18A inputs the CT image to each of lesion detection AIs 18B to 18F. Each of lesion detection AIs 18B to 18F performs lesion detection processing on the CT image and outputs the results as second discrimination results.

[0056] In step S6, the processor 16B acquires and integrates the second discrimination results from the second discrimination unit 18A (process P6).

[0057] In step S7, processor 16B determines whether or not an abnormality (here, a lesion) exists in the CT image based on the combined second discrimination result. If an abnormality exists in the CT image (an example of the "second case", process P7), the process proceeds to step S8, and if an abnormality does not exist in the CT image (an example of the "third case", process P8), the process proceeds to step S9.

[0058] In step S8, the processor 16B causes the display 20B of the PACS viewer 20 to display the CT image containing an abnormality in display format B, and performs post-processing of processing format B (process P9). The post-processing of processing format B is an example of post-processing in the second case where the medical image is determined to be abnormal in the first determination and to contain an abnormality in the second determination. Furthermore, the processor 16B adds "type B" incidental information to the CT image and stores it in the image storage server 14, and ends the processing of this flowchart.

[0059] On the other hand, in step S9, the processor 16B causes the display 20B of the PACS viewer 20 to display the CT image containing no abnormalities in display format C, and performs post-processing of processing format C (process P10). The post-processing of processing format C is an example of post-processing in the third case where the medical image is determined to be abnormal in the first determination and to be free of abnormalities in the second determination. Furthermore, the processor 16B assigns "type C" incidental information to the CT image and stores it in the image storage server 14, and ends the processing of this flowchart.

[0060] Processor 16B causes display 20B to display the determination results in different modes for display mode A, display mode B, and display mode C. That is, processor 16B causes display 20B to display the determination results in display mode A in a mode different from the modes for display mode B and display mode C. Processor 16B may also cause display 20B to display the determination results in different modes for display mode B and display mode C.

[0061] For example, display mode A indicates that there is a high possibility that there is no abnormality, allowing the doctor to skip confirmation and automatically reporting to the patient that there is no abnormality. Display mode B visually displays the name and area of ​​any detected lesions, just like regular CAD. Furthermore, display mode C notifies the doctor that no lesions were detected, but that the image was not clearly normal.

[0062] FIG. 4 is a diagram showing display form A. As shown in FIG. 4, in display form A, a CT image I1 is displayed on the display 20B. Also, in display form A, an explanatory text T1 for the CT image I1 is displayed in the area to the right of the CT image I1. Here, explanatory text T1 saying "Determined normal by CAD" is displayed on the display 20B. Note that, because the CT image I1 is likely not abnormal, it is possible to simply display a message indicating normality without displaying the CT image I1, thereby skipping the doctor's confirmation.

[0063] Fig. 5 is a diagram showing display form B. As shown in Fig. 5, in display form B, a CT image I2 is displayed on display 20B, and a marker M1 surrounding the lesion area in CT image I2 is superimposed on the CT image I2. Also, in display form B, an explanatory text T2 for CT image I2, which is an explanatory text T2 regarding the lesion area surrounded by marker M1, is displayed in the area to the right of CT image I2. Here, the lesion area has been detected by lesion detection AI 18B manufactured by Company A that detects disease α (lung cancer), and explanatory text T2 stating "Detected by lung cancer detection CAD manufactured by Company A" is displayed on display 20B.

[0064] FIG. 6 is a diagram showing display form C. As shown in FIG. 6, in display form C, a CT image I3 is displayed on display 20B, and a marker M2 that surrounds the entire CT image I3 is superimposed on the CT image I3. Also, in display form C, an explanatory text T3 for the CT image I3, which is an explanatory text T3 related to marker M2, is displayed in the area to the right of the CT image I3. Here, explanatory text T3 stating "Cannot be determined as normal by normality determination AI. However, no abnormalities have been reported for any CAD" is displayed on display 20B. In this way, by using a display form different from display form A and display form B, it is possible to correct judgments that rely too heavily on the lesion detection AI.

[0065] The processor 16B performs post-processing in different modes for the processing mode A, the processing mode B, and the processing mode C.

[0066] For example, in processing mode A, a flag is set that allows doctors to easily check the images, and no flag is set in processing modes B and C. In addition, the display order of the interpretation and examination list for doctors to check medical images may be changed so that CT images in processing mode B and C are given priority over CT images in processing mode A.

[0067] In the case of health checkups, it is rare for diseases that require immediate hospitalization or immediate treatment to be found, so there are cases where the results are not notified on the spot, or the subject is initially told that their results are normal, and then it is possible to separately inform them that abnormalities have been found. Therefore, in processing mode A, it is possible to have the subject report "no abnormalities" on the spot, and then later check whether there are actually no abnormalities. For example, in processing mode A, a doctor may perform a simple check at the end of the day. If an abnormality is found in the simple check, the subject can be contacted separately.

[0068] The processor 16B may perform post-processing in different ways for the processing type B and the processing type C.

[0069] As described above, according to the medical image diagnosis method, the first discrimination unit 16A can determine whether a medical image is normal. Furthermore, if the first discrimination unit 16A determines that the medical image is not normal, the second discrimination unit 18A can determine whether an abnormality exists in the medical image. Therefore, the burden on a doctor when performing image diagnosis on a large number of medical images can be reduced.

[0070] 〔others〕 For a medical image determined to be normal by the first discrimination unit 16A, a third discrimination process may be performed by the second discrimination unit 18A, for example, after the hospital closes (an example of a different timing), in which the presence or absence of an abnormality is determined from the medical image. The third discrimination process may be performed when the medical image is determined to be normal in the first discrimination with a probability lower than a predetermined first threshold. The third discrimination process is also an example of post-processing in the first case in which the medical image is determined to be normal in the first discrimination.

[0071] If the third discrimination process determines that an abnormality exists, the normality discrimination AI 16D may be retrained using a training data set of the CT image and the abnormal label, thereby allowing the normality discrimination AI 16D to determine that the CT image is abnormal.

[0072] Furthermore, if it is determined in the third discrimination process that an abnormality exists, the sensitivity of the normality discrimination AI16D may be changed to a value at which the CT image is considered to be abnormal.

[0073] For a medical image that has been determined to have no abnormality by the second discrimination unit 18A, a fourth discrimination process may be performed again by the second discrimination unit 18A to discriminate the presence or absence of an abnormality from the medical image. The fourth discrimination process may be performed with a higher sensitivity than the second discrimination process (with a lower specificity than the second discrimination process), i.e., with a relatively low threshold, until it is determined that an abnormality is present. The fourth discrimination process is also an example of post-processing in the third case where the medical image is determined to be abnormal in the first discrimination and to have no abnormality in the second discrimination.

[0074] In this way, the fourth discrimination process extracts lesions using evaluation criteria that allow for the misclassification of non-pathological conditions as pathological, and reports this to the doctor. In this case, if the display format is set to display format D, it is desirable to increase the sensitivity in display format D to present to the doctor the fact that a pathological condition has been extracted.

[0075] Fig. 7 is a diagram showing display form D. As shown in Fig. 7, in display form D, a CT image I3 is displayed on display 20B, a marker M2 surrounding the entire CT image I3 is superimposed on the CT image I3, and a marker M3 surrounding the lesion area of ​​the CT image I3 is further superimposed on the CT image I3. Unlike marker M1 in display form B, marker M3 is displayed with a dashed line to indicate that it is a lesion detected with increased sensitivity.

[0076] In addition, in display form D, an explanatory text T4 for the CT image I2, which is an explanatory text T4 regarding the marker M3, is displayed in the area to the right of the CT image I2. Here, the explanatory text T4 reads, "The normality determination CAD did not determine normality, so the sensitivity was increased and CAD processing was performed again, and lung cancer was detected by the detection CAD manufactured by company A." is displayed on display 20B.

[0077] 6, a slider bar SB for setting the sensitivity of the second discrimination unit 18A may be displayed in display form C. When the user increases the sensitivity by operating the slider bar SB using the input device 20A, the third discrimination process may be performed at the set sensitivity, and the display may transition to display form D shown in FIG.

[0078] For medical images that are determined to contain an abnormality by the second discrimination unit 18A, the above-described third and fourth discrimination processes are not performed, and the medical image and the discrimination result are associated with each other as usual and stored in the image storage server 14. This is also an example of post-processing in the second case where the medical image is determined to be abnormal in the first discrimination and to contain an abnormality in the second discrimination.

[0079] Although the present embodiment describes processing of a CT image of the lungs, the present invention is not limited to this. For example, for a CT image including the liver, each lesion detection AI of the second discrimination unit 18A may extract liver cancer, polycystic cysts, cirrhosis, and fatty liver, and the above-mentioned processes may be performed. Furthermore, the above-mentioned processes may be performed for medical images including other organs.

[0080] The technical scope of the present invention is not limited to the scope described in the above embodiments. The configurations and the like in each embodiment can be appropriately combined with each other within the scope that does not deviate from the spirit of the present invention. [Explanation of symbols]

[0081] 10...Medical image diagnostic system 12...Modality 14...Image storage server 16...Results integrated CAD processing server 16A...First discrimination section 16B...processor 16C...Memory 16D…Normal judgment AI 18...CAD processing servers of various companies 18A...Second discrimination section 18B… Lesion detection AI made by company A 18C... Lesion detection AI made by company A 18D... Lesion detection AI made by company A 18E…Lesion detection AI made by Company B 18F… Lesion detection AI made by Company C 20...PACS viewer 20A...input device 20B...Display I1...CT image I2...CT image I3...CT image M1...marker M2...marker M3...marker P1 to P10: Medical image diagnosis processes SB...Slider bar S1 to S9: Each step of medical image diagnosis T1...Description T2…Description T3…Description T4…Description

Claims

1. at least one processor; at least one memory storing instructions for execution by said at least one processor; Equipped with The at least one processor performing a first determination of whether a medical image obtained by imaging a subject is an image of a healthy subject; When the medical image is determined not to be an image of a healthy subject in the first determination, a second determination is performed to determine whether or not at least one of a disease, illness, and lesion is present in the medical image; displaying the discrimination result of the medical image on a display in different modes depending on whether the medical image is discriminated to be an image of a healthy subject in the first discrimination or whether the medical image is discriminated to not be an image of a healthy subject; Medical imaging diagnostic system.

2. The at least one processor When the medical image is determined to be an image of a healthy subject in the first determination, a third determination is performed to determine whether or not at least one of a disease, a disorder, and a lesion is present in the medical image. The medical image diagnostic system according to claim 1 .

3. The at least one processor The third determination is performed at a timing different from that of the second determination. The medical image diagnostic system according to claim 2 .

4. The at least one processor performing the first determination using a probability that the medical image is an image of a healthy subject; performing the third discrimination when the medical image is discriminated to be an image of a healthy subject with a probability lower than a first threshold in the first discrimination; The medical image diagnostic system according to claim 2 .

5. The at least one processor performing the first determination using a first trained model that, when the medical image is input, outputs whether the medical image is an image of a healthy subject; retraining the first trained model using medical images that have been determined to have at least one of a disease, a disorder, and a lesion in the third determination; The medical image diagnostic system according to claim 2 .

6. The first trained model is a trained model trained using a training dataset of a combination of a medical image having at least one of a disease, a medical condition, and a lesion, and a medical image not having at least one of a disease, a medical condition, and a lesion, and a label indicating the presence or absence of at least one of a disease, a medical condition, and a lesion. The medical image diagnostic system according to claim 5 .

7. The at least one processor When it is determined in the second determination that at least one of the disease, illness, and lesion is not present, a fourth determination is performed to determine the presence or absence of at least one of the disease, illness, and lesion from the medical image; In the fourth determination, the presence or absence of at least one of a disease, a disorder, and a lesion is determined from the medical image with a sensitivity relatively higher than that of the second determination. The medical image diagnostic system according to claim 1 .

8. The at least one processor performing the second discrimination using a second trained model that, when the medical image is input, outputs at least one of a disease, a disorder, and a lesion in the medical image; The medical image diagnostic system according to claim 1 .

9. The at least one processor a first case in which the medical image is determined to be an image of a healthy person in the first discrimination, a second case in which the medical image is determined to be not an image of a healthy person in the first discrimination and is determined to have at least one of the disease, illness, and lesion in the second discrimination, and a third case in which the medical image is determined to be not an image of a healthy person in the first discrimination and is determined to have at least one of the disease, illness, and lesion in the second discrimination, displaying the discrimination results of the medical image on a display in different modes for the first case, the second case, and the third case. The medical image diagnostic system according to claim 1 .

10. The at least one processor displaying the discrimination result of the medical image on the display in a manner different from that of the second case and the third case; The medical image diagnostic system according to claim 9 .

11. The at least one processor performing different post-processing on the medical image in the first case, the second case, and the third case, respectively; The medical image diagnostic system according to claim 9 .

12. The at least one processor performing the first discrimination and the second discrimination for each organ of the subject from the medical image; The medical image diagnostic system according to any one of claims 1 to 11.

13. 1. A method of operating a medical imaging diagnostic system comprising at least one processor, comprising: the at least one processor: performing a first determination of whether a medical image obtained by imaging a subject is an image of a healthy subject; When the medical image is determined not to be an image of a healthy subject in the first determination, a second determination is performed to determine whether or not at least one of a disease, illness, and lesion is present in the medical image; displaying the discrimination result of the medical image on a display in different modes depending on whether the medical image is discriminated to be an image of a healthy subject in the first discrimination or whether the medical image is discriminated to not be an image of a healthy subject; A method for operating a medical imaging system.

14. A program that causes a computer to execute the method for operating a medical image diagnostic system according to claim 13.

Images