Medical image diagnostic system, medical image diagnostic system evaluation method and program

Abstract

Provided are a medical image diagnosis system, a medical image diagnosis system evaluation method, and a program with which it is possible to establish the contribution of each of a plurality of discrimination results in an integrated discrimination result obtained by integrating the plurality of discrimination results. A first discrimination result (24A) is acquired from a first discrimination device (16A), a second discrimination result (24B) is acquired from a second discrimination device (16B) for carrying out a second discrimination performed on the same site and the same lesion as the first discrimination device, an integrated discrimination result (26) that integrates the first discrimination result and the second discrimination result is derived, and the degree of contribution of the first discrimination device to the integrated discrimination result and the degree of contribution of the second discrimination device to the integrated discrimination result are derived on the basis of the integrated discrimination result and a result of a definitive diagnosis by a physician.

Description

[Technical Field]

[0001] The present invention relates to a medical image diagnostic system, a method for evaluating a medical image diagnostic system, and a program. [Background technology]

[0002] Medical image diagnosis support systems that apply AI (Artificial Intelligence) to detect and diagnose abnormal areas in medical images are known. Medical image diagnosis support systems are referred to as CAD, an abbreviation of Computer Aided Diagnosis. It is expected that multiple CADs will be able to be used for the same area and the same lesion.

[0003] For example, in an image diagnosis platform that employs multiple CADs provided by multiple companies, a medical image is subjected to discrimination processing using each of the multiple CADs, and a user such as a doctor can refer to the discrimination results of each of the multiple CADs.

[0004] Patent Document 1 describes a composite image diagnosis support system that acquires diagnostic results output from multiple computer-aided diagnosis devices, determines the final result as a system by applying specified criteria, and presents the final result to an operator such as a doctor. In this document, the computer-aided diagnosis devices are referred to as CAD (Computer Aided Diagnosis). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2006-167289 Summary of the Invention [Problem to be solved by the invention]

[0006] However, even when the medical images are of the same site and the same lesion, it is cumbersome for doctors to distinguish and diagnose medical images that combine different CAD processing results.

[0007] The system described in Patent Document 1 aims to automatically construct a new database by sorting the images from which the final results are derived into a disease-bearing image database and a disease-free image database. The system described in Patent Document 1 does not focus on the above-mentioned problems, and the document does not disclose any components that solve the above-mentioned problems.

[0008] The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a medical image diagnostic system, a method for evaluating a medical image diagnostic system, and a program for evaluating the system, which are capable of grasping the contribution of each discrimination result with respect to an integrated discrimination result obtained by integrating a plurality of discrimination results. [Means for solving the problem]

[0009] The medical image diagnostic system according to the present disclosure includes a processor and a memory storing one or more instructions to be executed by the processor, wherein the processor acquires a first discrimination result from a first discrimination device that performs a first discrimination on a medical image targeting a predetermined region and a predetermined lesion, acquires a second discrimination result from a second discrimination device that performs a second discrimination on the medical image targeting the same predetermined region and predetermined lesion as the first discrimination device, derives an integrated discrimination result by integrating the first discrimination result and the second discrimination result, and derives the contribution of the first discrimination device to the integrated discrimination result and the contribution of the second discrimination device to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result by a doctor.

[0010] According to the medical image diagnostic system of the present disclosure, the contribution of the first discrimination result to the integrated discrimination result and the contribution of the second discrimination result to the integrated discrimination result are derived based on the integrated discrimination result, thereby making it possible to grasp the contributions of the first discrimination device and the second discrimination device to the integrated discrimination result obtained by integrating a plurality of discrimination results.

[0011] In another aspect of the medical image diagnostic system, a processor derives an integrated discrimination result using an integrated discrimination device, which is a trained learning model that has been trained using pairs of medical images and correct images in which lesions have been detected from the medical images as training data.

[0012] According to this aspect, it is possible to improve the discrimination accuracy of the integrated discrimination result.

[0013] The ground truth image may be a lesion mask image that represents a lesion in a medical image.

[0014] In another aspect of the medical image diagnostic system, the integrated classification device is configured to use a trained image data set that is trained using a first supervised answer data set output from the first classification device when a medical image is input to the first classification device and a second supervised answer data set output from the second classification device when a medical image is input to the second classification device as training data. study The model is used.

[0015] According to this aspect, it is possible to improve the discrimination accuracy of the integrated discrimination result.

[0016] In another aspect of the medical image diagnostic system, the processor derives a first score representing the contribution of the first discriminator to the integrated discrimination result, and derives a second score representing the contribution of the second discriminator to the integrated discrimination result.

[0017] According to this aspect, the contributions of the first and second discriminator devices to the integrated discrimination result can be evaluated based on the contributions quantified as scores.

[0018] In a medical image diagnostic system according to another aspect, the processor evaluates the contribution of the first discrimination device and the second discrimination device to the integrated discrimination result for each predetermined region.

[0019] According to this aspect, the characteristics of the first discriminator and the characteristics of the second discriminator can be evaluated for each region.

[0020] In a medical image diagnostic system according to another aspect, the processor evaluates the contribution of the first and second classification devices to an integrated classification result for each of the regions into which the predetermined part is subdivided.

[0021] According to this aspect, it is possible to perform a contribution evaluation on the integrated discrimination result that reflects the characteristics of each of the subdivided regions of the parts in the first discriminant model and the second discriminant model.

[0022] In a medical image diagnostic system according to another aspect, the processor causes at least one of the first discrimination result and the second discrimination result to be displayed on a display device.

[0023] According to this aspect, a user such as a doctor can grasp at least one of the first discrimination result and the second discrimination result.

[0024] In a medical image diagnostic system according to another aspect, a processor acquires input information representing a doctor's definitive diagnosis, which is input using an input device.

[0025] According to this aspect, a user such as a doctor can input the doctor's definitive diagnosis using the input device.

[0026] A medical image diagnostic system according to another aspect includes an image storage device in which medical images are stored, and a processor acquires the medical images from the image storage device.

[0027] According to this aspect, medical images can be acquired from an image storage device.

[0028] A medical image diagnostic system according to another aspect includes a first discrimination device and a second discrimination device.

[0029] According to this aspect, a medical image diagnostic system can be configured that includes the first discrimination device and the second discrimination device.

[0030] A medical image diagnostic system evaluation method according to the present disclosure is a medical image diagnostic system evaluation method in which a computer acquires a first discrimination result obtained by performing a first discrimination on a medical image targeting a predetermined region and a predetermined lesion, acquires a second discrimination result obtained by performing a second discrimination on the medical image targeting the same predetermined region and predetermined lesion as the first discrimination, derives an integrated discrimination result by integrating the first discrimination result and the second discrimination result, and derives a contribution of the first discrimination to the integrated discrimination result and a contribution of the second discrimination to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result by a doctor.

[0031] According to the medical image diagnostic system evaluation method of the present disclosure, it is possible to obtain the same effects as the medical image diagnostic system of the present disclosure. The constituent elements of the medical image diagnostic system of other aspects can be applied to the constituent elements of the medical image diagnostic system evaluation method of other aspects.

[0032] A program according to the present disclosure causes a computer to acquire a first discrimination result obtained by performing a first discrimination on a medical image targeting a specified area and a specified lesion, to acquire a second discrimination result obtained by performing a second discrimination on the medical image targeting a specified area and a specified lesion that is identical to the first discrimination, to derive an integrated discrimination result by integrating the first discrimination result and the second discrimination result, and to derive the contribution of the first discrimination to the integrated discrimination result and the contribution of the second discrimination to the integrated discrimination result based on the integrated discrimination result and a doctor's definitive diagnosis result.

[0033] According to the program of the present disclosure, it is possible to obtain the same effects as those of the medical image diagnostic system of the present disclosure. The components of the medical image diagnostic system of other aspects may be applied to the components of the program of other aspects. [Effects of the Invention]

[0034] According to the present invention, the contribution of the first discrimination result to the integrated discrimination result and the contribution of the second discrimination result to the integrated discrimination result are derived based on the integrated discrimination result, thereby making it possible to grasp the contributions of the first discrimination device and the second discrimination device to the integrated discrimination result obtained by integrating a plurality of discrimination results. [Brief explanation of the drawings]

[0035] [Figure 1] FIG. 1 is a block diagram of a medical image diagnostic system according to an embodiment. [Figure 2] Figure 2 is a schematic diagram of the learning process applied to the result-integrated CAD processing server. [Figure 3] FIG. 3 is a flowchart showing the procedure of the medical image diagnostic system evaluation method according to the embodiment. [Figure 4] FIG. 4 is a table showing an example of the evaluation results of the CAD processing server. [Figure 5] FIG. 5 is a schematic diagram showing an example of the configuration of the evaluation results of the CAD processing server. [Figure 6] FIG. 6 is a schematic diagram showing another example of the configuration of the CAD processing server. [Figure 7] FIG. 7 is a diagram showing an example of the discrimination result screen. [Figure 8] FIG. 8 is a diagram showing another example of the discrimination result screen. DETAILED DESCRIPTION OF THE INVENTION

[0036] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings. In this specification, the same components are designated by the same reference numerals, and redundant explanations will be omitted where appropriate.

[0037] [Configuration example of medical image diagnostic system] Figure 1 is a block diagram of a medical image diagnostic system according to an embodiment. The medical image diagnostic system 10 shown in the figure includes a modality 12, an image storage server 14, multiple CAD processing servers 16, a result integration CAD processing server 18, and a PACS viewer 20. CAD is an abbreviation for Computer-Aided Diagnosis. PACS is an abbreviation for Picture Archiving and Communication System.

[0038] In the medical image diagnostic system 10, the modality 12, the image storage server 14, the multiple CAD processing servers 16, the result integration CAD processing server 18, and the PACS viewer 20 are each connected to each other via a communication network such as the Internet so that they can send and receive data to and from each other.

[0039] In this embodiment, a collection of multiple devices is called a system, but the term system can include the concept of a device. That is, the terms system and device can be read interchangeably.

[0040] The modality 12 is an imaging device that captures an image of a target area of ​​a subject and generates a medical image 22. Examples of the modality 12 include an X-ray imaging device, a CT device, an MRI device, a PET device, an ultrasound device, and a CR device using a flat panel X-ray detector. Here, the term medical image is synonymous with medical image.

[0041] CT is an abbreviation for Computed Tomography, MRI is an abbreviation for Magnetic Resonance Imaging, PET is an abbreviation for Positron Emission Tomography, and CR is an abbreviation for Computed Radiography.

[0042] The image storage server 14 is a server that manages medical images 22 captured using the modality 12. A computer equipped with a large-capacity storage device is used as the image storage server 14. A program that provides the functions of a data storage system is installed in the computer. The image storage server 14 acquires medical images 22 captured using the modality 12 and stores them in the large-capacity storage device. Note that the term "program" is synonymous with software. The image storage server 14 described in the embodiment is an example of an image storage device in which medical images are stored.

[0043] The format of the medical image 22 can use the DICOM standard. DICOM tag information defined in the DICOM standard may be added to the medical image 22. The term "image" can mean not only the image itself, such as a photograph, but also image data, which is a signal representing an image. DICOM is an abbreviation for Digital Imaging and Communications in Medicine.

[0044] 1 illustrates a CAD processing server 16A manufactured by company A, a CAD processing server 16B manufactured by company B, and a CAD processing server 16C manufactured by company C, all of which are manufactured by three different companies. There is no limit to the number of CAD processing servers 16, as long as there is more than one.

[0045] The multiple CAD processing servers 16 perform anomaly detection processing for each part of the medical image 22 acquired from the image storage server 14, and determine whether or not there is an abnormality in the medical image. Each of the multiple CAD processing servers 16 can perform anomaly detection processing for a specific part and a specific lesion using the same medical image 22. That is, each of the multiple CAD processing servers 16 can perform anomaly detection processing for the same part and the same lesion based on the same medical image 22. Examples of parts include organs, bones, muscles, ligaments, nerves, blood vessels, etc. Examples of abnormalities include diseases, illnesses, and lesions. The determination results 24 from each of the multiple CAD processing servers 16 are associated with the medical image 22 to be processed, and the results are transmitted to the integrated CAD processing server 18.

[0046] The term "detection" may include the concept of "extraction," and the term "discrimination" may include concepts such as "identification," "recognition," "inference," "estimation," and "detection."

[0047] FIG. 1 shows an example in which a discrimination result 24A is output from a CAD processing server 16A manufactured by company A, a discrimination result 24B is output from a CAD processing server 16B manufactured by company B, and a discrimination result 24C is output from a CAD processing server 16C manufactured by company C.

[0048] The discrimination result 24 output from the CAD processing server 16 may be a binary image in which pixels detected as abnormal have a pixel value of 1 and pixels not detected as abnormal have a pixel value of 0. Fig. 1 shows examples of discrimination result 24A, discrimination result 24B, and discrimination result 24C in which a binary image is used.

[0049] 1 includes one or more processors and one or more memories. In the CAD processing server 16, the processor executes instructions included in a program stored in the memory to classify the medical image 22 and output the classification result 24.

[0050] The CAD processing server 16A manufactured by company A is equipped with a CAD engine 28A manufactured by company A. The CAD engine 28A manufactured by company A performs an abnormality detection process on the medical image 22 in the CAD processing server 16A manufactured by company A, and outputs a discrimination result 24A.

[0051] The company B CAD processing server 16B is equipped with a company B CAD engine 28B. The company B CAD engine 28B performs an abnormality detection process on the medical image 22 in the company B CAD processing server 16B and outputs a discrimination result 24B.

[0052] The company C CAD processing server 16C is equipped with a company C CAD engine 28C. The company C CAD engine 28C performs anomaly detection processing on the medical image 22 in the company C CAD processing server 16C and outputs a discrimination result 24C. The company A CAD engine 28A, the company B CAD engine 28B, and the company C CAD engine 28C may use a trained learning model such as a convolutional neural network called a CNN (Convolutional Neural Network).

[0053] The result integrated CAD processing server 18 acquires each discrimination result 24 for one medical image 22 transmitted from each of the multiple CAD processing servers 16, and generates an integrated discrimination result 26 based on each discrimination result 24 for one medical image 22. The integrated discrimination result 26 is associated with the medical image 22 to be processed, transmitted to the image storage server 14, and stored in the image storage server 14.

[0054] A trained learning model is used in the result integrated CAD processing server 18. An example of the learning model applied to the result integrated CAD processing server 18 is a convolutional neural network.

[0055] A convolutional neural network may have a configuration in which some of its multiple hidden layers are a combination of a convolutional layer and a pooling layer. The number of hidden layers constituting the convolutional neural network, the processing content of each layer, and the arrangement order of each layer are not limited, and a structure consisting of various combinations may be adopted.

[0056] The convolutional layer performs convolution operations using filters on nodes within a local region in the previous layer to obtain a feature map. The convolutional layer is responsible for feature extraction, extracting the characteristic grayscale structures represented by the filters from the image.

[0057] The pooling layer performs pooling processing to aggregate local regions of the feature map output from the convolutional layer using representative values, and reduces the feature map output from the convolutional layer to generate a new feature map with lower resolution.

[0058] The pooling layer provides robustness to the target features extracted using the convolutional layer so that they are not affected by positional variations. In other words, the pooling layer reduces the sensitivity of the target features to positional variations.

[0059] In addition to the convolutional layer and the pooling layer, the convolutional neural network may include one or more layers of at least one of the normalization layer and the fully connected layer. Each intermediate layer may include an activation function as needed.

[0060] The normalization layer performs a process of normalizing the gray-scale structure of an image. For example, the normalization layer normalizes the local contrast of at least one of the output of the convolution layer and the output of the pooling layer.

[0061] A fully connected layer is a layer that connects all nodes between adjacent layers. A fully connected layer can be placed near the output layer. For example, a fully connected layer connects a feature map, from which features have been extracted through a convolutional layer and a pooling layer, to a single node and outputs feature variables using an activation function. Generally, in a convolutional neural network, one or more fully connected layers are placed between the last pooling layer and the output layer. The output layer performs class classification using a softmax function or the like based on the output from the fully connected layer.

[0062] The result integrated CAD processing server 18 evaluates the CAD processing performance, such as the characteristics and features of CAD processing, for each of the multiple CAD processing servers 16 based on the integrated judgment result 26. The result integrated CAD processing server 18 stores the CAD processing performance evaluation results for each of the multiple CAD processing servers 16.

[0063] Furthermore, the result integration CAD processing server 18 acquires definitive diagnosis information representing the doctor's definitive diagnosis result, and evaluates the contribution of CAD processing to the integrated discrimination result 26 based on the integrated discrimination result 26 and the doctor's definitive diagnosis result. Note that the definitive diagnosis information described in the embodiment is an example of input information representing the doctor's definitive diagnosis result.

[0064] The result integrated CAD processing server 18 stores the results of the contribution evaluation of the CAD processing to the integrated discrimination result 26 of each of the multiple CAD processing servers 16. The performance evaluation and contribution evaluation for the multiple CAD processing servers 16 will be described in detail later.

[0065] The result integrated CAD processing server 18 is a computer. The computer may be a personal computer or a workstation. The result integrated CAD processing server 18 includes one or more processors 18A and one or more memories 18B. The processor 18A executes instructions stored in the memory 18B.

[0066] Examples of the hardware structure of the processor 18A include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a PLD (Programmable Logic Device), and an ASIC (Application Specific Integrated Circuit). A CPU is a general-purpose processor that executes programs and functions as various functional units. A GPU is a processor specialized for image processing.

[0067] A PLD is a processor whose electrical circuit configuration can be changed after the device is manufactured. An example of a PLD is an FPGA (Field Programmable Gate Array). An ASIC is a processor with dedicated electrical circuitry designed specifically to perform a specific task.

[0068] A 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. Examples of combinations of various processors include a combination of one or more FPGAs and one or more CPUs, and a combination of one or more FPGAs and one or more GPUs. Another example of a combination of various processors is a combination of one or more CPUs and one or more GPUs.

[0069] A single processor may be used to configure multiple functional units. An example of using a single processor to configure multiple functional units is a configuration in which a single processor is configured by applying a combination of one or more CPUs and software, such as an SoC (System On a Chip), which is typified by a computer such as a client or server, and this processor operates as multiple functional units.

[0070] Another example of using one processor to configure multiple functional units is to use a processor that uses one IC chip to realize the functions of an entire system including multiple functional units. Note that IC is an abbreviation for Integrated Circuit.

[0071] In this way, the various functional units are configured as hardware structures using one or more of the various processors described above.More specifically, the hardware structures of the various processors described above are electric circuits (circuitry) that combine circuit elements such as semiconductor elements.

[0072] The memory 18B stores instructions to be executed by the processor 18A. The memory 18B may include a random access memory (RAM) or a read only memory (ROM).

[0073] The processor 18A uses RAM as a working area, executes software using various programs and parameters including a medical image processing program stored in ROM, and executes various processes of the result-integrated CAD processing server 18 using parameters stored in ROM, etc.

[0074] The result integrated CAD processing server 18 may have the functions of multiple CAD processing servers 16. Hardware such as the processor 18A of the result integrated CAD processing server 18 may be used to execute programs used by the multiple CAD processing servers 16. The result integrated CAD processing server 18 may have the hardware and software of the multiple CAD processing servers 16 built in.

[0075] The PACS viewer 20 is a terminal device used by a user such as a doctor. A known image viewer for image interpretation is used as the PACS viewer 20. The PACS viewer 20 may be a personal computer, a workstation, or a tablet terminal.

[0076] The PACS viewer 20 includes an input device 20A and a display device 20B. Examples of the input device 20A include 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.

[0077] The display device 20B functions as a GUI (Graphical User Interface) that displays a screen required for operations using the input device 20A. The display device 20B also displays medical images captured by the modality 12. Furthermore, the display device 20B displays the integrated discrimination result 26 as the CAD result 29.

[0078] That is, the display device 20B receives a display signal representing the CAD result 29 and displays the CAD result 29. The CAD result 29 includes the integrated discrimination result 26. The CAD result 29 may also include the discrimination result 24 for each CAD processing server 16.

[0079] The PACS viewer 20 may be a touch panel display in which the input device 20A and the display device 20B are integrated.

[0080] The CAD processing server 16A manufactured by Company A, the CAD processing server 16B manufactured by Company B, and the CAD processing server 16C manufactured by Company C described in the embodiment are examples of a first discrimination device that performs the first discrimination and an example of a second discrimination device that performs the second discrimination. In other words, the CAD processing server 16A manufactured by Company A, the CAD processing server 16B manufactured by Company B, and the CAD processing server 16C manufactured by Company C are each either a first discrimination device or a second discrimination device. For example, if the CAD processing server 16A manufactured by Company A is the first discrimination device, the CAD processing server 16B manufactured by Company B and the CAD processing server 16C are the second discrimination device. The discrimination results 24A, 24B, and 24C described in the embodiment are examples of a first discrimination result and an example of a second discrimination result. In other words, the discrimination results 24A, 24B, and 24C are each either a first discrimination result or a second discrimination result. For example, if the discrimination result 24A is the first discrimination result, the discrimination results 24B and 24C are the second discrimination results. The result integration CAD processing server 18 described in the embodiment is an example of an integration determination device.

[0081] [Learning applied to the result-integrated CAD processing server] 2 is a schematic diagram of the learning applied to the result integrated CAD processing server 18. In the learning of the result integrated CAD processing server 18, a pair of a training medical image 30 and a correct mask image 32 is used as learning data.

[0082] As a result, the training medical images 30 used as training data for the integrated CAD processing server 18 are medical images captured using the same modality as the medical images used as training data for the multiple CAD processing servers 16.

[0083] Furthermore, the training medical images 30 are medical images of the same site and the same lesion as the medical images used as training data for the multiple CAD processing servers 16. On the other hand, the training medical images 30 are medical images that are not used as training data for the multiple CAD processing servers 16.

[0084] 2 illustrates a lung CT image as a training medical image 30, and an example of a lung nodule region mask image as a correct mask image 32. The lung nodule region mask image can be generated by performing mask processing on a lung nodule region extracted from a lung CT image.

[0085] The following procedure is used for training the result-integrated CAD processing server 18. A training medical image 30 serving as training data is input to each of a CAD engine 28A made by company A, a CAD engine 28B made by company B, and a CAD engine 28C made by company C.

[0086] Each of company A's CAD engine 28A, company B's CAD engine 28B, and company C's CAD engine 28C processes the training medical image 30 and acquires a lung tumor labeling image as a provisional supervised image 34. Specifically, company A's CAD engine 28A acquires provisional supervised image 34A, company B's CAD engine 28B acquires provisional supervised image 34B, and company C's CAD engine 28C acquires provisional supervised image 34C.

[0087] FIG. 2 illustrates a provisional correct image 34A in which two lung nodules 35 are detected, a provisional correct image 34B in which three lung nodules 35 are detected, and a provisional correct image 34C in which one lung nodule 35 is detected.

[0088] A set of the provisional correct answer image 34A, the provisional correct answer image 34B, and the provisional correct answer image 34C is input to the result integration CNN 18C. When the set of the provisional correct answer image 34A, the provisional correct answer image 34B, and the provisional correct answer image 34C is input, the result integration CNN 18C performs learning to output the correct mask image 32.

[0089] During training of the result integration CAD processing server 18, a set of provisional supervised images 34A, 34B, and 34C may be input to the result integration CNN 18C instead of the training medical image 30. The provisional supervised images 34A, 34B, and 34C described in the embodiment are examples of first supervised data and also examples of second supervised data. For example, if the provisional supervised image 34A is the first supervised data, each of the provisional supervised images 34B and 34C is the second supervised data. The term "data" includes the concepts of signal and information.

[0090] [Medical image diagnostic system evaluation method according to the embodiment] The medical image diagnostic system 10 evaluates the characteristics of each of the multiple CAD processing servers 16 based on the integrated discrimination result 26. Furthermore, the medical image diagnostic system 10 evaluates the contribution of each of the multiple CAD processing servers 16 to the integrated discrimination result 26 based on the integrated discrimination result 26 and the doctor's definitive diagnosis result.

[0091] Examples of the characteristics of the CAD processing server 16 include areas that the server is good at, lesions that the server is good at, and types of medical images that the server is good at. The areas that the server is good at may be subdivided regions. The types of medical images can be understood as the types of modalities that generate the medical images. The types of medical images include, for example, MRI images and CT images.

[0092] The medical image diagnostic system 10 can evaluate the characteristics of the CAD processing server 16 and re-learn the result integration CNN 18C based on the evaluation results. For example, if the CAD processing server 16A manufactured by Company A excels at discriminating the right lung, the result integration CAD processing server 18 can re-learn the result integration CNN 18C by actively adopting the discrimination result 24A of the CAD processing server 16A manufactured by Company A in the discrimination process for the right lung. In other words, when performing discrimination process for the right lung, the result integration CAD processing server 18 weights the discrimination result of the CAD processing server 16 that excels at discriminating the right lung and re-learns the result integration CNN 18C. That is, the result integration CAD processing server 18 weights the discrimination result of the CAD processing server 16 that excels at discriminating the part targeted by the discrimination process and re-learns the result integration CNN 18C, depending on the part targeted by the discrimination process.

[0093] The contribution of the CAD processing server 16, which indicates its contribution to the integrated discrimination result 26, can also be considered as the necessity of each CAD processing server 16. The contribution of the CAD processing server 16 can be derived based on the number of cases adopted in the integrated discrimination result 26 and the number of cases adopted in the doctor's definitive diagnosis.

[0094] This allows the user of the medical image diagnostic system 10 to know, based on the contribution of each CAD processing server 16, which CAD processing server 16 has a relatively low contribution to improving the discrimination accuracy of the integrated discrimination result 26.

[0095] Furthermore, the user of the medical image diagnostic system 10 may stop charging for a CAD processing server 16 that is making a relatively low contribution, remove it from the lineup, etc. Also, the user may urge the company that manages the CAD processing server 16 to improve the CAD processing server 16, etc.

[0096] That is, the medical image diagnostic system 10 performs performance evaluation for all of the multiple CAD processing servers 16, and calculates and stores the contribution to the integrated discrimination result 26. With each discrimination process, the user of the medical image diagnostic system 10 can grasp the redundancy of the discrimination result 24 and the performance of the CAD processing servers 16, and can narrow down the use of the multiple CAD processing servers 16 to one or two CAD processing servers 16.

[0097] 3 is a flowchart showing the procedure of the medical image diagnostic system evaluation method according to the embodiment. In the discrimination result acquisition step S10, the result integration CAD processing server 18 shown in FIG. 1 acquires a set of discrimination results 24 for one medical image from each of the multiple CAD processing servers 16, and stores the acquired set of discrimination results 24.

[0098] In the integrated discrimination result deriving step S12, the result integrated CAD processing server 18 derives an integrated discrimination result 26 based on the set of discrimination results 24 acquired in the discrimination result acquisition step S10, and stores the integrated discrimination result 26.

[0099] Specifically, in the integrated discrimination result derivation step S12, a set of the discrimination result 24A, the discrimination result 24B, and the discrimination result 24C is input to the trained result integration CNN 18C, and the integrated discrimination result 26 is output from the result integration CNN 18C.

[0100] In the integrated discrimination result display step S14, the result integrated CAD processing server 18 displays the integrated discrimination result 26 on the display device 20B.

[0101] In the definitive diagnosis information acquisition step S16, the result integrated CAD processing server 18 acquires definitive diagnosis information including the doctor's definitive diagnosis input using the input device 20A. The result integrated CAD processing server 18 associates the acquired definitive diagnosis information with the integrated discrimination result 26 and stores it.

[0102] In the contribution degree deriving step S18, the contribution degree for each CAD processing server 16 is derived based on the integrated discrimination result 26 and the definitive diagnosis information. The contribution degree for each CAD processing server 16 may be a score representing the contribution of the discrimination result 24 to the integrated discrimination result 26.

[0103] In the contribution degree storage step S20, the result integrated CAD processing server 18 stores the contribution degree for each CAD processing server 16 derived in the contribution degree derivation step S18.

[0104] 3 may be performed when a new integrated discrimination result 26 of the medical image 22 is acquired, or may be performed based on a user input signal. The contribution of each CAD processing server 16 may be updated every time the procedure shown in FIG. 3 is performed.

[0105] [Detailed explanation of CAD processing server evaluation] Fig. 4 is a table showing an example of the evaluation results of the CAD processing servers. The figure shows the evaluation results of the CAD processing server 16A made by company A, the CAD processing server 16B made by company B, and the CAD processing server 16C made by company C shown in Fig. 1. The numbers in the column marked "times" in the table shown in Fig. 4 indicate the number of times the discrimination process was performed. The number of times the discrimination process was performed shown in the table of Fig. 4 can be understood as the identification number of the medical image to be processed.

[0106] In the table shown in Figure 4, "Detected" indicates a case where an abnormality candidate such as a lesion candidate is detected from the medical image 22. "Not Detected" indicates a case where an abnormality candidate is not detected from the medical image 22. For example, the CAD processing server 16A manufactured by company A and the CAD processing server 16C manufactured by company C detected an abnormality candidate in all of the first through fourth runs. On the other hand, the CAD processing server 16B manufactured by company B detected an abnormality candidate in the second run, but did not detect an abnormality candidate in the first, third, or fourth runs.

[0107] In the table shown in Fig. 4, "adopted" indicates a case where the discrimination result was adopted in the doctor's definitive diagnosis. "Not adopted" indicates a case where the discrimination result was not adopted in the doctor's definitive diagnosis. For example, in the case of the CAD processing server 16A manufactured by company A, the discrimination results were adopted in the doctor's definitive diagnosis in all of the first to fourth times.

[0108] On the other hand, although the CAD processing server 16B manufactured by Company B detected a possible abnormality the second time, the discrimination result was not adopted in the doctor's definitive diagnosis, and the discrimination result of the CAD processing server 16A manufactured by Company A and the discrimination result of the CAD processing server 16C manufactured by Company C were adopted in the doctor's definitive diagnosis.

[0109] On the other hand, the CAD processing server 16C manufactured by Company C detected an abnormality candidate the first time, but the discrimination result was not adopted in the doctor's definitive diagnosis, and the discrimination result of the CAD processing server 16A manufactured by Company A was adopted in the doctor's definitive diagnosis. The discrimination results of the CAD processing server 16C manufactured by Company C were adopted in the doctor's definitive diagnosis from the second to fourth times.

[0110] Based on the evaluation results of the CAD processing servers shown in Figure 4, the user can understand that the CAD processing server 16A made by company A and the CAD processing server 16C made by company C have similar characteristics. Specifically, in the example shown in Figure 4, the CAD processing server 16A made by company A and the CAD processing server 16C made by company C detect abnormality candidates for the same medical image 22. In other words, servers that similarly detect abnormality candidates for the same medical image 22 can be understood to have similar characteristics. Furthermore, the CAD processing server 16A made by company A has a higher number of discrimination results that have been adopted in a doctor's definitive diagnosis than the CAD processing server 16C made by company C. In other words, the doctor's Confirmed The more times a discrimination result is used in the diagnosis, the higher the contribution, and the fewer times a discrimination result is used, the lower the contribution. This allows the user to consider stopping use of the CAD processing server 16C made by company C, which has a relatively low contribution, between the CAD processing server 16A made by company A and the CAD processing server 16C made by company C, which have similar characteristics. The user may also consider stopping use of the CAD processing server 16B made by company B, which has the lowest contribution.

[0111] The result integrated CAD processing server 18 may calculate a score representing the contribution of each CAD processing server 16 based on the evaluation results of the CAD processing servers 16 shown in FIG. 4. In other words, the result integrated CAD processing server 18 may calculate a score representing the contribution based on the detection characteristics (performance) of the abnormality candidate and the result of adoption of the definitive diagnosis. Specifically, a high score may be assigned when an abnormality candidate is detected and adopted in the definitive diagnosis, a medium score may be assigned when an abnormality candidate is detected but not adopted in the definitive diagnosis, and a low score may be assigned when no abnormality candidate is detected. For example, a positive score of 1 point may be assigned when an abnormality candidate is detected and adopted in the doctor's definitive diagnosis, a negative score of 0.5 point may be assigned when an abnormality candidate is detected but not adopted in the doctor's definitive diagnosis, and a negative score of 2 points may be assigned when no abnormality candidate is detected. The result integrated CAD processing server 18 may calculate an overall evaluation score for each CAD processing server 16 by adding up the above scores.

[0112] The result integration CAD processing server 18 may issue a warning when the ratio of the number of times an abnormal candidate is detected in the number of times of discrimination processing is equal to or less than a specified value, or may issue a warning when the ratio of the number of times an abnormal candidate is adopted in the number of times of discrimination processing is equal to or less than a specified value. ratio of If the value is less than the specified value, a warning may be issued.

[0113] The score for each CAD processing server 16 described in the embodiment is an example of the first score and an example of the second score. In other words, the score of each CAD processing server 16A, 16B, and 16C is the first score or the second score.

[0114] FIG. 5 is a schematic diagram showing an example of the configuration of the evaluation results of the CAD processing server. FIG. 5 shows the evaluation results of the CAD processing server 16 derived for each body part. Specifically, FIG. 5 shows the evaluation results of the CAD processing server 16 derived for the lungs, the evaluation results of the CAD processing server 16 derived for the heart, and the evaluation results of the CAD processing server 16 derived for the bronchi. The result integration CAD processing server 18 evaluates the CAD processing server 16 for each body part and derives the evaluation results of the CAD processing server 16 for each body part. The medical image diagnostic system 10 can derive the contribution of each CAD processing server 16 for each body part. Furthermore, based on the evaluation results of the CAD processing server 16 for each body part, the medical image diagnostic system 10 can evaluate the performance of the CAD processing server 16 for each body part, such as its specialty body part.

[0115] Fig. 6 is a schematic diagram showing another example of the configuration of the evaluation results of the CAD processing server. Fig. 6 shows the evaluation results of the CAD processing server 16 derived for each region into which the body part is subdivided. Fig. 6 shows the right lung, left lung, lung area S1, and lung area S2 as examples of regions.

[0116] For example, for pulmonary nodules, the CAD processing server 16 may be evaluated separately depending on the nature and location of the pulmonary nodule, and whether or not to use the CAD processing server 16 may be optimized depending on the nature and location of the pulmonary nodule. In other words, multiple CADs may be used for each region corresponding to the type of lesion and the location of the lesion. process The servers 16A, 16B, and 16C may be evaluated in advance, and the appropriate CAD processing server 16A, 16B, or 16C may be used depending on the type and location of the lesion.

[0117] [Example of display screen configuration] Fig. 7 is a diagram showing an example of a discrimination result screen. Discrimination result screen 100 shown in Fig. 7 is one mode for notifying integrated discrimination result 26 shown in Fig. 1, and is displayed using display device 20B shown in Fig. 1. On discrimination result screen 100, a CT image I1 is displayed in first area 102, and markers M1 and M2 surrounding a lesion area detected from CT image I1 are superimposed on CT image I1.

[0118] In the discrimination result screen 100, a second area 104 arranged to the right of the CT image I1 displays an explanatory text T1 for the CT image I1, which is an explanatory text T1 regarding the lesion area surrounded by the marker M1. In Fig. 7, the explanatory text T1 for the CT image I1 displays text information indicating that the CT image I1 was detected using the CAD processing server 16A manufactured by company A.

[0119] Fig. 8 is a diagram showing another example of the discrimination result screen. The discrimination result screen 120 shown in Fig. 8 is displayed using the display device 20B shown in Fig. 1. The discrimination result screen 120 is used when no lesion area is detected from the CT image I1.

[0120] On the discrimination result screen 120, the CT image I1 is displayed in the first area 102. On the other hand, on the discrimination result screen 120, the markers M1 and M2, etc. shown in FIG.

[0121] On the discrimination result screen 120, a description T2 for the CT image I1 is displayed in the second area 104. In Fig. 8, text information indicating that no abnormality was detected using the CAD processing server 16B manufactured by company B is displayed as the description T2 for the CT image I1. The description T2 shown in Fig. 8 may be displayed together with the description T1 shown in Fig. 7.

[0122] The result integrated CAD processing server 18 can notify the discrimination results 24 of the multiple CAD processing servers 16 shown in FIG. 1 on a discrimination result display screen that displays an integrated discrimination result 26, as shown in FIGS.

[0123] The result integrated CAD processing server 18 may selectively switch between displaying and hiding the explanatory text T1 and the like for the CT image I1. For example, the server 18 may selectively switch between displaying and hiding the explanatory text T1 and the like in accordance with selection information input by the user using the input device 20A shown in FIG.

[0124] [Effects of the embodiment] The medical image diagnostic system, medical image diagnostic system evaluation method, and program according to the embodiment can achieve the following advantageous effects.

[0125] [1] In a medical image diagnostic system 10 that derives an integrated discrimination result 26 based on discrimination results 24 of a plurality of CAD processing servers 16 for the same site and the same lesion, a contribution indicating the contribution of the discrimination result 24 to the integrated discrimination result 26 is derived. This makes it possible to identify CAD processing servers 16 that have a relatively low contribution to improving the accuracy of the integrated discrimination result 26.

[0126] [2] A contribution indicating the contribution of the discrimination result 24 to the integrated discrimination result 26 is derived for each part and for each region obtained by subdividing the part. This makes it possible to grasp the strengths and weaknesses of each CAD processing server 16 for each part and region. It also makes it possible to grasp the necessity of the CAD processing server 16 for each part and region.

[0127] [3] Based on whether or not the doctor has adopted the definitive diagnosis, a contribution indicating the contribution of the discrimination result 24 to the integrated discrimination result 26 is derived. This allows the CAD processing server 16 to be evaluated based on the doctor's definitive diagnosis.

[0128] [Modification of medical image diagnostic system] The components constituting the medical image diagnostic system shown in Fig. 1 can be integrated or separated as appropriate. For example, the result integration CAD processing server 18 and the like may be configured using multiple computers.

[0129] For example, part or all of the image storage server 14 and part or all of the result integration CAD processing server 18 may be configured using a single computer.

[0130] 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]

[0131] 10 Medical imaging diagnostic systems 12 Modalities 14 Image storage server 16 CAD processing servers 16A CAD processing server manufactured by company A 16B CAD processing server manufactured by Company B 16C CAD processing server made by C company 18 Result integrated CAD processing server 18A processor 18B memory 20 PACS Viewer 20A input device 20B Display device 22 Medical Imaging 24 Judgment results 24A Judgment result 24B Judgment result 24C discrimination result 26 Integrated discrimination results 28A CAD engine manufactured by A company 28B B company CAD engine 28C C company CAD engine 29 CAD results 30 Educational Medical Images 32 Correct mask images 34 Tentative correct image 34A Tentative correct image 34B Tentative correct image 34C Tentative correct image 35 Lung mass 100 screens 102 First area 104 Second area 120 screens I1CT images M1 marker M2 marker T1 Description T2 Description S10~S20 Each step of medical image diagnosis

Claims

1. a processor; a memory for storing one or more instructions for execution by the processor; Equipped with The processor: acquiring a first discrimination result from a first discrimination device that performs a first discrimination targeting a predetermined region and a predetermined lesion on the medical image; obtaining a second discrimination result from a second discrimination device that performs a second discrimination on the medical image, the second discrimination being the same as that performed by the first discrimination device, and targeting the predetermined region and the predetermined lesion; deriving an integrated discrimination result by integrating the first discrimination result and the second discrimination result; a medical image diagnostic system that derives a degree of contribution of the first discriminator to the integrated discrimination result and a degree of contribution of the second discriminator to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result of a doctor;

2. A processor; a memory for storing one or more instructions for execution by the processor; Equipped with The processor: acquiring a first discrimination result from a first discrimination device that performs a first discrimination targeting a predetermined region and a predetermined lesion on the medical image; obtaining a second discrimination result from a second discrimination device that performs a second discrimination on the medical image, the second discrimination being the same as that performed by the first discrimination device, and targeting the predetermined region and the predetermined lesion; deriving an integrated discrimination result by integrating the first discrimination result and the second discrimination result; deriving a contribution degree of the first discriminator to the integrated discrimination result and a contribution degree of the second discriminator to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result of a doctor; The processor derives the integrated classification result using an integrated classification device that is a trained model that has been trained using pairs of the medical image and a correct image in which a lesion is detected from the medical image as training data. Medical imaging diagnostic system.

3. the integrated classification device uses the trained learning model trained using, as training data, a set of the correct image, first correct data output from the first classification device when the medical image is input to the first classification device, and second correct data output from the second classification device when the medical image is input to the second classification device; The medical image diagnostic system according to claim 2 .

4. The processor: deriving a first score representing a contribution of the first discriminator to the integrated discrimination result; deriving a second score representing a contribution of the second discriminator to the integrated discrimination result; The medical image diagnostic system according to claim 1 .

5. the processor evaluates the contribution of the first discriminator and the second discriminator to the integrated discrimination result for each of the predetermined regions. The medical image diagnostic system according to any one of claims 1 to 4.

6. the processor evaluates the contribution of the first discriminator and the second discriminator to the integrated discrimination result for each of the subdivided regions of the predetermined part. The medical image diagnostic system according to any one of claims 1 to 5.

7. the processor causes a display device to display at least one of the first determination result and the second determination result. The medical image diagnostic system according to any one of claims 1 to 6.

8. the processor acquires input information representing the doctor's definitive diagnosis result input using an input device; The medical image diagnostic system according to any one of claims 1 to 7.

9. an image storage device in which the medical images are stored; the processor acquires the medical image from the image storage device; The medical image diagnostic system according to any one of claims 1 to 8.

10. The first discrimination device and the second discrimination device are provided. The medical image diagnostic system according to any one of claims 1 to 9.

11. The processor, deriving a contribution of the first discriminator to the integrated discrimination result based on a detection characteristic of the first discriminator for the abnormality candidate and an adoption result of the first discrimination result in a definitive diagnosis; deriving a contribution of the second discriminator to the integrated discrimination result based on the detection characteristics of the second discriminator for the abnormality candidate and the result of adopting the second discrimination result as a definitive diagnosis; The medical image diagnostic system according to any one of claims 1 to 10.

12. The processor, displaying the discrimination result of the medical image and information about the discrimination device that derived the discrimination result; The medical image diagnostic system according to any one of claims 1 to 11.

13. The computer a first discrimination result obtained by performing a first discrimination targeting a predetermined region and a predetermined lesion on the medical image; a second discrimination result obtained by performing a second discrimination, which is the same as the first discrimination and targets the predetermined region and the predetermined lesion, on the medical image; deriving an integrated discrimination result by integrating the first discrimination result and the second discrimination result; deriving a contribution degree of the first discrimination to the integrated discrimination result and a contribution degree of the second discrimination to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result of a doctor; Medical imaging diagnostic system evaluation method.

14. A computer comprising: a first discrimination result obtained by performing a first discrimination targeting a predetermined region and a predetermined lesion on the medical image; a second discrimination result obtained by performing a second discrimination, which is the same as the first discrimination and targets the predetermined region and the predetermined lesion, on the medical image; deriving an integrated discrimination result by integrating the first discrimination result and the second discrimination result; deriving a contribution degree of the first discrimination to the integrated discrimination result and a contribution degree of the second discrimination to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result of a doctor; deriving the integrated discrimination result using an integrated discrimination device that is a trained model that has been trained using pairs of the medical image and a correct image in which a lesion has been detected from the medical image as training data; Medical imaging diagnostic system evaluation method.

15. deriving a contribution of the first discrimination device to the integrated discrimination result based on the detection characteristics of the abnormality candidate of the first discrimination device that performs the first discrimination and the result of adopting the first discrimination result to a definitive diagnosis; deriving a contribution of the second discriminator to the integrated discrimination result based on a detection characteristic of the second discriminator that performs the second discrimination and a result of adopting the second discrimination result as a definitive diagnosis; The method for evaluating a medical image diagnostic system according to claim 13 or 14.

16. Displaying the discrimination result of the medical image and information about the discrimination device that derived the discrimination result. The method for evaluating a medical image diagnostic system according to any one of claims 13 to 15.

17. On the computer, performing a first discrimination targeting a predetermined region and a predetermined lesion on the medical image, and acquiring a first discrimination result; performing, on the medical image, a second discrimination that is the same as the first discrimination and targets the predetermined region and the predetermined lesion, to obtain a second discrimination result; deriving an integrated discrimination result by integrating the first discrimination result and the second discrimination result; deriving a contribution degree of the first discrimination to the integrated discrimination result and a contribution degree of the second discrimination to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result of a doctor; program.

18. A computer comprising: performing a first discrimination targeting a predetermined region and a predetermined lesion on the medical image, and acquiring a first discrimination result; performing, on the medical image, a second discrimination that is the same as the first discrimination and targets the predetermined region and the predetermined lesion, to obtain a second discrimination result; deriving an integrated discrimination result by integrating the first discrimination result and the second discrimination result; deriving a contribution degree of the first discrimination to the integrated discrimination result and a contribution degree of the second discrimination to the integrated discrimination result based on the integrated discrimination result and a definitive diagnosis result of a doctor; deriving the integrated discrimination result using an integrated discrimination device that is a trained model that has been trained using pairs of the medical image and a correct image in which a lesion has been detected from the medical image as training data; program.

19. The computer: deriving a contribution of the first discriminator to the integrated discrimination result based on a detection characteristic of an abnormality candidate of a first discriminator that performs the first discrimination and a result of adopting the first discrimination result into a definitive diagnosis; deriving a contribution of the second discriminator to the integrated discrimination result based on a detection characteristic of an abnormality candidate of the second discriminator that performs the second discrimination and a result of adopting the second discrimination result as a definitive diagnosis; 19. The program according to claim 17 or 18.

20. The computer: displaying the discrimination result of the medical image and information about the discrimination device that derived the discrimination result; 20. The program according to any one of claims 17 to 19.

Images