JP7442964B2 - Medical information collection system and medical information collection device - Google Patents

Description

Embodiments of the present invention relate to a medical information gathering system (hereinafter referred to as an information gathering system) and a medical information gathering apparatus (hereinafter referred to as an information gathering apparatus) .

Image processing techniques using machine learning are known. In machine learning, for example, a trained model configured by a multilayer neural network is used. This technology requires a large amount of training data containing a variety of images with different characteristics in order to improve the accuracy of the trained model.

When training a trained model, if there is a large amount of monotonous training data with a high degree of similarity, the accuracy of discrimination may decrease when rare-case images are input.

US Publication No. 2016/0361025 US Patent No. 9208672 US Publication No. 2017/0177798

The problem to be solved by the present invention is to collect useful learning data.

The information collection system of the embodiment includes a first storage section, an acquisition section, a determination section, and a second storage section. The first storage unit includes a first trained model that includes one or more intermediate layers that perform calculation processing on input information, and an output layer that outputs diagnostic results based on the calculation results of the intermediate layers. Information regarding the calculation results output by the intermediate layer is stored. The acquisition unit acquires a calculation result output from the intermediate layer in response to medical information being input to the input layer of the first trained model. The determination unit determines whether the calculation result acquired by the acquisition unit satisfies a condition with information regarding the calculation result stored in the first storage unit. The second storage unit stores information regarding the medical information corresponding to the calculation result determined to satisfy the condition.

FIG. 1 is a diagram showing an example of the configuration of an information collection system according to the first embodiment. FIG. 2 is a diagram for explaining the flow of learning processing according to the first embodiment. FIG. 3 is a diagram illustrating an example of distribution of calculation results according to the first embodiment. FIG. 4 is a diagram illustrating an example of an activation function according to the first embodiment. FIG. 5 is a diagram illustrating an example of functional blocks of the information collection system according to the first embodiment. FIG. 6 is a diagram illustrating an example of the flow of learning processing according to the first embodiment. FIG. 7 is a diagram illustrating an example of the flow of determination processing according to the first embodiment. FIG. 8 is a diagram illustrating an example of calculation result information according to the first embodiment. FIG. 9 is a diagram showing an example of statistical data according to the first embodiment. FIG. 10 is a diagram illustrating an example of useful data according to the first embodiment. FIG. 11 is a diagram showing an example of model data according to the first embodiment. FIG. 12 is a diagram illustrating an example of the flow of information collection processing according to the first embodiment. FIG. 13 is a diagram showing an example of a reception screen according to the first embodiment. FIG. 14 is a flowchart illustrating an example of the acquisition process according to the first embodiment. FIG. 15 is a flowchart illustrating an example of the determination process according to the first embodiment. FIG. 16 is a flowchart illustrating an example of the reception process according to the first embodiment. FIG. 17 is a diagram illustrating an example of functional blocks of the information collection system according to the second embodiment. FIG. 18 is a diagram illustrating an example of useful data according to the second embodiment. FIG. 19 is a diagram illustrating an example of functional blocks of the information collection system according to the third embodiment. FIG. 20 is a diagram illustrating an example of distribution of calculation results according to the fourth embodiment.

Embodiments of an information collection system and an information collection device will be described in detail below with reference to the drawings. The information collection system according to the present embodiment is useful for improving the accuracy of a trained model, such as information that corresponds to rare cases among medical information input to a trained model, for example in CAD (Computer Aided Diagnosis). Collect information. The learned model in this embodiment can be configured by, for example, a multilayer neural network. Note that in the following, intermediate calculation results of the learned model may be referred to as "intermediate calculation results" or simply "calculation results." Furthermore, hereinafter, "information collection" may be simply referred to as "collection."

For example, when a rare case is input to a trained model, unusual values are often output in the calculation results in the intermediate layer, and as a result, the discrimination results may become invalid. For example, for input information that should be determined as "lung cancer," a determination result may be output in which the accuracy of "pulmonary nodule" is higher than the accuracy of "lung cancer." In this case, the output discrimination result may not be useful as supporting information for a doctor to diagnose whether it is "lung cancer" or "pulmonary nodule." Note that in the following, the term "rare case" is not limited to rare cases, but also includes medical information that includes parameters that are considered to be outside the learned range in the trained model, even if the case is not rare.

For example, when training a trained model, if there is a lot of monotonous training data with a high degree of similarity, the robustness against data that corresponds to rare cases will be impaired.For example, images that include parameters that are outside the trained range The accuracy of discrimination against data may become low. Therefore, in order to improve the accuracy of the learned model, it is useful to prepare training data in which image data corresponding to rare cases and correct answer data are associated with each other.

Note that, for example, in a multilayer neural network, it is often not possible to clearly specify which parameters of the learning data are reflected in the trained model, that is, to what extent the trained range is. Therefore, it may be difficult to determine whether medical information input to a trained model corresponds to a rare case in the trained model. Therefore, in this embodiment, a configuration will be described in which medical image data corresponding to rare cases is collected using the calculation results output by the intermediate layer in the trained model. Note that in the following, determining that medical image data "corresponds to a rare case" is an example of medical information "satisfying a condition."

(First embodiment)
First, a first embodiment will be described. In the first embodiment, an information collection system 1 including an information collection device 10 will be described as an example.

FIG. 1 is a diagram showing an example of the configuration of an information collection system according to the first embodiment. As shown in FIG. 1, the information collection system 1 according to the present embodiment includes an information collection device 10, a group of medical image diagnostic devices 20, an image storage device 21, a learning device 30, and a terminal 40.

As shown in FIG. 1, the information collection device 10, the medical image diagnostic device group 20, the learning device 30, and the terminal 40 are connected to each other so as to be able to communicate with each other via the network NW. Note that although FIG. 1 shows an information collection system 1 that includes one information collection device 10, one image storage device 21, one learning device 30, and one terminal 40, the number of each device is arbitrary. In addition, the information collection system 1 directly or indirectly communicates with each server device (not shown) included in other systems in the hospital such as a hospital accounting system (not shown) and an electronic medical record system (not shown). , may be communicably connected to each other. Further, in the following, a device described as a single computer device may be realized by a group of servers (cloud) connected via a network NW.

The information collecting device 10 collects useful information for improving the accuracy of the learned model, such as information corresponding to rare cases, from among the medical information input to the learned model. The information gathering device 10 determines whether the medical information is useful information, for example, using calculation results such as feature amounts output by the intermediate layer of the trained model during the discrimination processing by the learning device 30. . The information collecting device 10 collects information related to medical information determined to be useful information as needed. Further, the information collecting device 10 outputs information related to the collected medical information at an arbitrary timing, and receives input of, for example, correct data from a CAD administrator (not shown). For example, the information gathering device 10 is realized by computer equipment such as a workstation.

The medical image diagnostic device group 20 is a plurality of devices (modalities) that collect medical information from a subject. Note that medical information that is processed as image data is also referred to as medical image data. As shown in FIG. 1, the information collection system 1 in this embodiment includes a plurality of modalities XXX, XXY, and YYY as a medical image diagnostic apparatus group 20. In the following, when a plurality of modalities XXX, XXY, and YYY are expressed without distinction, they may be simply expressed as "modality."

For example, the medical image diagnostic device group 20 collects medical image data from a subject and transmits the collected medical image data to the learning device 30. An example of the medical image diagnostic device group 20 is an X-ray diagnostic device, and an X-ray CT (Computed Tomography) device, a SPECT (Single Photon Emission Computed Tomography) device, or an integrated SPECT device and an X-ray CT device. Other medical image diagnostic devices such as SPECT-CT devices and ultrasound diagnostic devices may also be used. Further, in FIG. 1, the information collection system 1 has three modalities as the medical image diagnostic apparatus group 20, but the number of modalities that the information collection system 1 according to the present embodiment has is not limited to this.

The image storage device 21 is a database (DB) or the like that stores medical image data collected by the medical image diagnostic device group 20 via the network NW. For example, in FIG. 1, an image storage device 21 acquires medical image data from a group of medical image diagnostic devices 20, and stores the acquired medical image data in a memory provided within or outside the device. Further, the image storage device 21 may further store correct data including information such as a case corresponding to the medical image data. For example, the image storage device 21 is realized by computer equipment such as a server device.

The learning device 30 acquires medical image data from the medical image diagnostic device group 20 and the like via the network NW, and executes various processes using the acquired medical image data. For example, the learning device 30 acquires medical image data from the medical image diagnostic device group 20 or the image storage device 21 via the network NW. The learning device 30 inputs the acquired medical image data into a learned model, performs a discrimination process, and outputs a corresponding case and a reliability level indicating the probability of corresponding to the case. Further, the learning device 30 performs learning on the learned model using learning data that includes medical image data serving as input information to the learned model and correct answer data such as a case as a data set. For example, the learning device 30 is realized by computer equipment such as a workstation.

In this embodiment, the learning device 30 has, for example, three trained models 341, 342, and 343, as shown in FIG. 1, but the number of trained models according to this embodiment is not limited to this. Furthermore, the trained models 341, 342, and 343 may be implemented in separate learning devices. The trained models 341, 342, and 343 are trained models used in CAD used in hospitals, respectively.

The terminal 40 is installed, for example, in a consulting room of a hospital, and outputs the determination results of the learning device 30 to a user such as a doctor. The terminal 40 is, for example, a desktop computer as shown in FIG. 1, but is not limited to this, and may be a portable computer such as a smartphone or a tablet.

Note that the information gathering device 10, the medical image diagnostic device group 20, the image storage device 21, the learning device 30, and the terminal 40 can be installed at any location as long as they can be connected via the network NW. For example, the learning device 30 may be installed in a hospital different from the hospital where the information gathering device 10 is installed. That is, the network NW may be configured as a local network closed within the hospital, or may be a network via the Internet.

Next, information collection processing in the information collection system 1 will be explained using FIG. 2. FIG. 2 is a diagram for explaining the flow of learning processing according to the first embodiment. As shown in FIG. 2, the learning device 30 receives input of medical image data 91 such as X-ray image data from the medical image diagnostic device group 20, for example. The learning device 30 inputs medical image data 91 to a learned model configured by a multilayer neural network. As shown in FIG. 2, the multilayer neural network according to this embodiment has a total of (L-2) layers including an input layer (l=1) and an intermediate layer (l=2, 3, ..., L-1). ), and an output layer (l=L).

The learning device 30 uses, for example, the pixel values of pixels forming the medical image data 91 as each component xp (p=1, 2, . . . , N) of the input signal x. In the intermediate layer following the input layer (l = 2, 3, ..., L-1 layer), by sequentially performing arithmetic processing on the input signal, each of the (L-2) intermediate layers The output z can be calculated. Furthermore, each intermediate layer has a plurality of filters. The output layer L can calculate the output yn (n=1, 2, . . . , M) by performing arithmetic processing on the output z of each layer. The output yn indicates, for example, the reliability for each case. That is, in the example shown in FIG. 2, the output yn outputs the reliability for each of M cases. Note that for the arithmetic processing in each layer, for example, a known neural network technique can be used, so a detailed explanation will be omitted.

The learning device 30 outputs information regarding the determination result based on the output yn in the output layer L to the terminal 40. For example, as shown in FIG. 2, the terminal 40 outputs information regarding the determination result output from the learning device 30. In the example shown in FIG. 2, a determination result is output in which the probability of lung nodule is 60% and the probability of lung cancer is 30%. Although FIG. 2 shows an example in which the terminal 40 displays the reliability of each case as information regarding the discrimination result, the terminal 40 is not limited to this, and the terminal 40 simply displays only the case with the highest reliability. You may also display other information such as. For example, the terminal 40 may display "There is a possibility of pulmonary nodule." Further, for example, the terminal 40 may display, in the medical image data 91, an area where a pulmonary nodule is suspected with 60% certainty.

In the present embodiment, the information gathering device 10 acquires the calculation results in the intermediate layer of the trained model of the learning device 30, and collects medical information that is recognized to correspond to rare cases. In the present embodiment, the information gathering device 10 acquires calculation results in, for example, a pooling layer among the intermediate layers of a trained model configured by a multilayer neural network, but is not limited to this, and is not limited to this. The calculation results for each layer such as the Linear Unit) layer may also be obtained.

In this embodiment, the information gathering device 10 sequentially acquires and stores the calculation results output in the intermediate layer, for example, during learning in the learning device 30 and during discrimination processing. Further, the information gathering device 10 determines whether the newly acquired calculation result is based on a rare case, for example, by comparing the newly acquired calculation result with the stored calculation result.

Regarding the criteria for determining whether an input image is a rare case, FIG. 3 is a diagram showing an example of the distribution of calculation results according to the first embodiment. In FIG. 3, the horizontal axis shows the value of the calculation result, and the vertical axis shows the frequency of appearance of the calculation result corresponding to the value. For example, as shown in Figure 3, when the frequency of occurrence of calculation results is normally distributed, the calculation results are within ±3σ (hereinafter, standard deviation may be referred to as "σ") from the average value μ. 97a is determined not to be based on a rare case. On the other hand, calculation results 97b and 97c that deviate from the range of ±3σ from the average value μ are determined to be based on rare cases. Note that the criterion for determining whether the calculation result is based on a rare case is merely an example, and may be determined based on other criteria. Furthermore, in a case where the frequency of appearance of the calculation result does not follow a normal distribution, it may be similarly determined whether the calculation result is based on a rare case based on the distance from the average value μ.

Features of calculation results based on rare cases are often not propagated to the output layer by the activation function. FIG. 4 is a diagram illustrating an example of an activation function according to the first embodiment. FIG. 4 shows an example of a graph of the ReLU function, which is often used in the middle layer of a trained model configured by a multilayer neural network. For example, even if the calculation result based on a rare case deviates from the average value μ of the calculation result by −3σ or more, the output will be 0 when input to the ReLU function as shown in FIG. Therefore, the characteristics when rare cases are input to the input layer are often not reflected in the discrimination results in the output layer, so it is difficult to determine whether input information is a rare case based on the discrimination results. difficult. Therefore, in this embodiment, a configuration will be described in which it is determined whether or not input information is a rare case using calculation results in the intermediate layer.

In such a configuration, the information collection device 10 in this embodiment acquires the calculation results output from the intermediate layer (l=2, 3, ..., L-1 layer) of the learned model of the learning device 30. do. Then, the information collecting device 10 determines whether the medical image data is useful as learning data based on the calculation result, and collects information regarding the useful medical image data.

FIG. 5 is a diagram illustrating an example of functional blocks of the information collection system according to the first embodiment. As shown in FIG. 5, the information collection system 1 includes an information collection device 10, a group of medical image diagnostic devices 20, an image storage device 21, a learning device 30, and a terminal 40. As shown in FIG. 5, the learning device 30 includes a communication I/F (Interface) 31, an input I/F 32, a display 33, a storage circuit 34, and a processing circuit 35. Note that each function of the processing circuit 35 of the learning device 30 may be implemented on the cloud, and the contents stored in the storage circuit 34 may be stored on the cloud.

The communication I/F 31 is connected to the processing circuit 35 and transmits and communicates various data with the information collection device 10, medical image diagnostic device group 20, image storage device 21, etc. connected via the network NW. control. For example, the communication I/F 31 is realized by a network card, a network adapter, a NIC (Network Interface Controller), or the like.

The input I/F 32 is connected to the processing circuit 35 , converts an input operation received from, for example, a CAD administrator into an electrical signal, and outputs the electrical signal to the processing circuit 35 . For example, the input I/F 32 is a switch button, a mouse, a keyboard, a touch panel, or the like.

The display 33 is connected to the processing circuit 35 and displays various information and various image data output from the processing circuit 35. For example, the display 33 is realized by a liquid crystal monitor, a CRT (cathode ray tube) monitor, a touch panel, or the like.

The storage circuit 34 is connected to the processing circuit 35 and stores various data. For example, the storage circuit 34 is realized by a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, a hard disk, an optical disk, or the like. In this embodiment, the storage circuit 34 stores a program for the circuit included in the learning device 30 to realize its function. Further, for example, the storage circuit 34 stores a plurality of trained models 341, 342, and 343, as shown in FIG. Note that the storage circuit 34 may be realized by a server group (cloud) connected to the learning device 30 via the network NW.

The trained models 341, 342, and 343 are trained models that are incorporated into the discrimination process executed by the processing circuit 35 to discriminate cases based on input medical image data. The trained models 341, 342, and 343, for example, as shown in FIG. ) consists of a multilayer neural network consisting of L layers. The learned models 341, 342, and 343 are generated by a model generation function 352, which will be described later.

Returning to FIG. 5, the processing circuit 35 executes a control function 351, a model generation function 352, a determination function 353, and an output function 354. Here, for example, each processing function of a control function 351, a model generation function 352, a discrimination function 353, and an output function 354, which are components of the processing circuit 35 shown in FIG. It is recorded in 34. The processing circuit 35 is a processor that reads each program from the storage circuit 34 and implements a function corresponding to each program by executing each read program. In other words, the processing circuit 35 in a state where each program is read has each function shown in the processing circuit 35 of FIG.

Note that all the processing functions of the control function 351, model generation function 352, discrimination function 353, and output function 354 may be recorded in the storage circuit 34 in the form of one program executable by a computer. For example, such a program is also referred to as a learning program. In this case, the processing circuit 35 reads the learning program from the storage circuit 34 and implements the control function 351, model generation function 352, discrimination function 353, and output function 354 corresponding to the learning program by executing the read learning program. .

A control function 351 in the processing circuit 35 controls operations on the learning device 30, various processes related to communication with other devices, and various processes related to data acquisition from other devices. For example, in response to a trigger to start the model generation process, the control function 351 acquires learning data including a set of medical image data and correct data from the image storage device 21, and outputs it to the model generation function 352. . Further, for example, the control function 351 acquires medical image data from, for example, the image storage device 21 and outputs it to the determination function 353 in response to a trigger for starting the determination process. Note that the trigger for starting the learning process and the discrimination process may be received from the terminal 40 or the like via the communication I/F 31, for example, by receiving an operation from the CAD administrator via the input I/F 32. .

The model generation function 352 executes processing related to generation of learned models 341, 342, and 343. For example, when the model generation function 352 first receives input of learning data from the control function 351, the model generation function 352 determines correct data such as cases from medical image data included in the learning data based on the learning data. , train the model. FIG. 6 is a diagram illustrating an example of the flow of learning processing according to the first embodiment.

As shown in FIG. 6, for example, the model generation function 352 adjusts model parameters by using medical image data included in the learning data as input data and correct data included in the learning data as output data. , generate a trained model that approximates a function that discriminates cases, etc. When the trained model to be generated is a trained model by a neural network, the parameter to be adjusted is, for example, at least one of weights and biases. The model generation function 352 then stores the generated trained models 341, 342, and 343 in the storage circuit 34. Furthermore, the learned models 341, 342, and 343 output calculation results in the intermediate layer. Note that, when new learning data is acquired, the model generation function 352 may perform relearning processing on the learned models 341, 342, and 343 in order to improve discrimination accuracy.

Note that the model generation function 352 may use medical image data included in the learning data as input side data, or may use data based on medical image data as input side data. For example, as shown in FIG. 6, the model generation function 352 executes preprocessing (for example, image processing using a spatial filter or convolution filter) on medical image data, and inputs the preprocessed image data. data may be used as input data.

The types of image processing filters include, for example, a moving average (smoothing) filter, a Gaussian filter, a median filter, a recursive filter, a bandpass filter, and the like. In addition, the parameters of the image processing filter include, for example, the strength of each filter, the vertical and horizontal sizes of the moving average (smoothing) filter, Gaussian filter, and median filter, the number of frames used in the recursive filter, and the Examples include weights, frequency settings in bandpass filters, etc.

The discrimination function 353 executes discrimination processing on input medical image data based on the learned model 341, 342, or 343. Specifically, the discrimination function 353 executes discrimination processing for discriminating cases, for example, by inputting preprocessed medical image data into the trained model 341, 342, or 343. Then, the determination function 353 outputs the determination result output from the trained model 341, 342, or 343 to the output function 354. FIG. 7 is a diagram illustrating an example of the flow of determination processing according to the first embodiment.

For example, the determination function 353 first reads out the learned model 341, 342, or 343 from the storage circuit 34. Next, as shown in FIG. 7, the discrimination function 353 inputs the medical image data to the learned model 341, 342, or 343, and executes discrimination processing to discriminate the case. The learned model 341, 342, or 343 outputs a discrimination result based on the input medical image data. Further, the trained model 341, 342, or 343 outputs the calculation result in the intermediate layer. Note that, as shown in FIG. 7, the determination function 353 may perform preprocessing on the medical image data and use the preprocessed image data as input side data, similarly to the learning process.

The output function 354 outputs the determination result by the determination function 353. Specifically, the output function 354 outputs the determination results output from the trained models 341, 342, and 343 to the terminal 40 shown in FIG. 5 via the communication I/F 31.

Next, the information gathering device 10 includes a communication I/F 11, an input I/F 12, a display 13, a storage circuit 14, and a processing circuit 15, as shown in FIG. Note that each function of the processing circuit 15 of the information gathering device 10 may be implemented on the cloud, and the contents stored in the storage circuit 14 may be stored on the cloud.

The communication I/F 11 is connected to the processing circuit 15 and controls transmission of various data and communication with the learning device 30 and the like connected via the network NW. For example, the communication I/F 11 is realized by a network card, a network adapter, or the like.

The input I/F 12 is connected to the processing circuit 15 , converts input operations received from the CAD administrator into electrical signals, and outputs the electrical signals to the processing circuit 15 . The input I/F 12 accepts input operations such as mode selection by a CAD administrator, for example. For example, the input I/F 12 is a switch button, a mouse, a keyboard, a touch panel, or the like.

The display 13 is connected to the processing circuit 15 and displays various information outputted from the processing circuit 15. The display 13 is realized by, for example, a liquid crystal monitor, a CRT monitor, a touch panel, or the like.

The storage circuit 14 is connected to the processing circuit 15 and stores various data. For example, the memory circuit 14 is realized by a semiconductor memory element such as a RAM or a flash memory, a hard disk, an optical disk, or the like. In this embodiment, the storage circuit 14 stores a program for the circuits included in the information collecting device 10 to realize their functions. Further, for example, the storage circuit 14 stores calculation result information 141, statistical data 142, useful data 143, and model data 144, as shown in FIG. Note that the storage circuit 14 may be realized by a server group (cloud) connected to the information gathering device 10 via the network NW. Note that the memory circuit 14 is an example of a first memory section.

The calculation result information 141 stores calculation results output from the intermediate layer of the learned model 341 in the learning process or discrimination process in the learning device 30. For example, the calculation result information 141 stores the calculation results output in the learning process shown in FIG. 6 and the calculation results output in the discrimination process shown in FIG. 7. The calculation result information 141 is stored, for example, by a control function 151, which will be described later. Note that the calculation result information 141 is an example of information stored in the first storage unit.

FIG. 8 is a diagram illustrating an example of calculation result information according to the first embodiment. For example, as shown in FIG. 8, the calculation result information 141 holds one table for each filter included in each intermediate layer of each trained model. The computation result information 141 shown in FIG. 8 stores, for example, the computation result output from the "2" filter of the "1" intermediate layer of the learned model with the model ID "M001".

For example, as shown in FIG. 8, the calculation result information 141 is information in which "acquisition date and time", "image ID", "processing category", and "calculation result" are associated with each other. “Acquisition date and time” indicates the date and time when the calculation result was acquired. "Image ID" indicates an identifier (IDentifier) that uniquely identifies the input medical image data when the calculation result is output. The "processing classification" indicates whether the calculation result is output during "learning" or "judgment". “Computation result” indicates the computation result output from the filter of the corresponding layer in the intermediate layer. The calculation results accumulated in the calculation result information 141 are used, for example, to generate statistical data to be described later. In Fig. 8, for example, in "during learning" using "image ID: XXX1", the output from the "2" filter of the "1" middle layer is output at "June 2, 2018, 15:20". It is stored that the calculated result is "0.604". In addition, in FIG. 8, for example, in the "discrimination time" for "image ID: XXY1", at "June 30, 2018 10:00", the "2" filter of the "1" middle layer outputs the It is stored that the calculated result is "0.317".

Note that the items shown in FIG. 8 are just examples, and the calculation result information 141 may further store other information such as "discrimination result", and some of the items shown in FIG. 8 may not be stored. It may be configured like this. Further, the calculation result information 141 may hold one table for each calculation result obtained in all filters of one layer, or may hold all calculation results in one table. The same applies to each table described below.

The statistical data 142 is statistical data such as the average value μ and standard deviation σ of the calculation results, which are calculated by statistically processing the calculation results stored in the calculation result information 141. The statistical data 142 is stored, for example, by the control function 151. Note that the statistical data 142 is an example of information stored in the first storage unit.

FIG. 9 is a diagram showing an example of statistical data according to the first embodiment. For example, the statistical data 142 holds one table for each trained model, as shown in FIG. The statistical data 142 shown in FIG. 9 stores, for example, data calculated by statistically processing calculation results output from each filter of each intermediate layer of a trained model with model ID "M001".

For example, as shown in FIG. 9, the statistical data 142 includes "middle layer No.", "filter No.", "number of records", "average value", "-3σ" and "+3σ". This is associated information. Here, "intermediate layer No." and "filter No." in FIG. 9 indicate information that uniquely identifies the layer and filter to which the calculation result is output. “Number of records” indicates the number of calculation results output from the filter in the layer and stored in the calculation result information 141. "Average value" indicates the average value μ of the calculation result. "-3σ" and "+3σ" each indicate a value that is plus or minus 3σ away from the average value μ of the calculation result.

In FIG. 9, "average value", "-3σ", and "+3σ" are used, for example, as threshold values for determining whether the obtained calculation result is based on a rare case. For example, if the calculation result deviates from the average value μ by "-3σ" or "+3σ", the determination function 152, which will be described later, determines that the calculation result is based on a rare case. In FIG. 9, for example, if the calculation result output from the "1" filter of the "1" hidden layer is smaller than "0.431" or larger than "0.674", the calculation result is due to a rare case. It is determined that it is a thing. Similarly, in FIG. 9, for example, if the calculation result output from the "2" filter of the "1" intermediate layer is smaller than "-0.932" or larger than "0.450", the calculation result is is determined to be due to a rare case.

The useful data 143 stores information regarding medical image data determined to be useful. The useful data 143 is stored, for example, by the determination function 152. Note that the useful data 143 is an example of information stored in the second storage unit.

FIG. 10 is a diagram illustrating an example of useful data according to the first embodiment. For example, as shown in FIG. 10, the useful data 143 includes "image ID", "modality ID", "patient ID", "model ID", "correct data", "doctor ID", This is information that is associated with the "relevant reason." Here, "image ID" in FIG. 10 indicates an identifier that uniquely identifies the input medical image data when the calculation result is output. "Modality ID" indicates information that uniquely identifies the modality that collected medical image data. “Patient ID” indicates information that uniquely identifies a patient, etc. for whom medical image data has been collected. “Model ID” indicates information that uniquely identifies a trained model used in the discrimination process of medical image data. "Correct data" and "doctor ID" respectively indicate correct data for medical image data and information that uniquely identifies the doctor who input the correct data. The “relevant reason” indicates the reason why the medical image data is determined to correspond to useful information.

For example, the table shown in FIG. 10 shows that the medical image data with image ID "XXX1" collected from the patient with patient ID "0001" by the modality with modality ID "XXX" is transferred to the trained model with model ID "M001". This indicates that it has been determined in the determination process that there is a "suspected rare case." Furthermore, the table shown in FIG. 10 indicates that the value of the n-th filter of the m-th layer of the learned model "M001" satisfies the condition in the discrimination process using the model ID "M001" for medical image data with the image ID "XXX1". shows. Note that the table shown in FIG. 10 includes only records related to calculation results that were determined to be "suspicious of rare cases," but is not limited to this; it also includes records related to calculation results that were not determined to be rare cases. May include. Note that the trained models with model IDs "M001", "M002", and "M003" correspond to, for example, trained models 341, 342, and 343 shown in FIG. 1, respectively. Further, the corresponding reason "corresponds to the target range" written in the record of the model ID "M003" will be explained in detail later.

The model data 144 stores information regarding a learned model, which is used, for example, in preprocessing to generate a learning model from medical image data. FIG. 11 is a diagram showing an example of model data according to the first embodiment. For example, as shown in FIG. 11, the model data 144 is information that associates "model ID", "input format", "filter type", and "corresponding modality". In FIG. 11, "Model ID" indicates information that uniquely identifies a trained model used for medical image data discrimination processing. "Input format" indicates information regarding the format, such as the size, of medical image data input to the model. “Filter type” indicates information regarding a filter used for preprocessing medical image data when inputting the medical image data to the learned model.

For example, the first record of the model data 144 shown in FIG. 11 indicates that the model with model ID "M001" applies "format A" and "filter A" and is compatible with all modalities. In addition, the third record of the model data 144 shown in FIG. 11 shows that the model with model ID "M003" applies "Format C" and "Filter C", and also supports the modalities of "CT" and "MR". Show that.

Returning to FIG. 5, the processing circuit 15 executes a control function 151, a determination function 152, an output function 153, and a reception function 154. Here, the control function 151 is an example of an acquisition unit. The determination function 152 is an example of a determination unit. The output function 153 is an example of an output unit. Further, the reception function 154 is an example of a reception unit. Here, for example, each processing function of a control function 151, a determination function 152, an output function 153, and a reception function 154, which are components of the processing circuit 15 shown in FIG. recorded in The processing circuit 15 is a processor that reads each program from the storage circuit 14 and executes each read program to implement a function corresponding to each program. In other words, the processing circuit 15 in a state where each program has been read has each function shown in the processing circuit 15 of FIG.

Note that all the processing functions of the control function 151, determination function 152, output function 153, and reception function 154 may be recorded in the storage circuit 14 in the form of one program executable by a computer. For example, such a program is also referred to as an information gathering program. In this case, the processing circuit 15 reads the information collection program from the storage circuit 14 and executes the read information collection program to perform the control function 151, determination function 152, output function 153, and reception function 154 corresponding to the information collection program. Realize.

A control function 151 in the processing circuit 15 controls operations on the information gathering device 10, various processes related to communication with other devices, and various processes related to data acquisition from other devices. For example, the control function 151 acquires the calculation result corresponding to the medical image data input to the learning device, for example, from the learning device 30 at any time, and outputs it to the determination function 152. Furthermore, the control function 151 performs statistical processing on the obtained calculation results and stores them in the statistical data 142.

Furthermore, the control function 151 obtains, for example, a patient ID of a patient whose medical image data input to the learning device has been collected. For example, when the information collection system 1 has a plurality of modalities XXX, XXY, and YYY as shown in FIG. 1, the control function 151 further acquires each modality ID. Further, the control function 151 further acquires the model ID of each learned model, for example, when the learning device 30 has a plurality of learned models 341, 342, and 343 as shown in FIG.

In this embodiment, the information gathering process is executed whenever the control function 151 detects the start of the learning process or the start of the discrimination process in the learning device 30, for example.

The determination function 152 determines whether the medical image data accompanying the calculation result is useful. For example, when the determination function 152 first receives the output of the calculation result from the control function 151, it refers to the statistical data 142 and determines whether the calculation result is in the range from "-3σ" to "+3σ" as shown in FIG. It is determined whether there is a deviation, that is, whether the calculation result is based on a rare case. FIG. 12 is a diagram illustrating an example of the flow of information collection processing according to the first embodiment. For example, as shown in FIG. 12, the determination function 152 compares the intermediate calculation result outputted from the control function 151 with the statistical data 142, and converts the results into useful data 143 as shown in FIG. 10 as a determination result. Store.

The determination function 152 identifies, as useful data, for example, input data in which the calculation result output from the intermediate layer of the learned model deviates from the range of "from -3σ to +3σ", that is, corresponds to a rare case. However, the embodiment is not limited to this, and the determination function 152 may, for example, determine that even if the calculation result is within the range of "-3σ to +3σ", it is considered useful for generating learning data. The configuration may be such that input data whose calculation result falls within the range is determined to be useful data. For example, when the number of samples of learning data whose calculation result falls within the range "from -σ to the average value μ" is small, the determination function 152 determines that data whose calculation result falls within the range is useful data. It may be configured such that For example, as shown in FIG. 10, the determination function 152 determines in the useful data 143 that the calculation result of the n-th filter of the m-th layer "corresponds to the target range" for the medical image data of the patient ID "0005". May be stored.

The output function 153 outputs information regarding the collected medical image data as useful information. The output function 153 outputs, to the display 13 of the information collecting device 10, a screen that accepts input of correct answer data for medical image data, for example. Note that although a case will be described below in which correct answer data is input from a CAD administrator who refers to the display 13, the output destination of the screen is not limited to this. For example, when a CAD administrator operates another computer such as the terminal 40, the output function 153 may output a screen to the computer through the communication I/F 11.

FIG. 13 is a diagram showing an example of a reception screen according to the first embodiment. As shown in FIG. 13, for example, a reception screen 80 displayed on the display 13 includes a display area 81 where information regarding medical image data is displayed, an input area 82 that accepts input of correct answer data, a decision button 83, and a medical image data 91. The display area 81 includes information for identifying medical image data, such as an image ID, modality ID, and patient ID. The medical image data 91 is medical image data of the image ID displayed in the display area 81. The input area 82 is an area for accepting input of correct data such as cases by, for example, a CAD administrator. Note that the correct answer data is an example of a determination result.

The reception function 154 receives information regarding correct answer data input on the reception screen 80. Upon detecting that the OK button 83 has been pressed on the reception screen 80, the reception function 154 acquires information regarding the case input in the input area 82. The reception function 154 stores the acquired information regarding the case as correct data in the useful data 143 in association with the image ID, modality ID, patient ID, model ID, etc. displayed on the reception screen 80. Further, the reception function 154 identifies the doctor who has inputted the information regarding the case, and stores the doctor ID in the useful data 143.

For example, as shown in FIG. 10, the reception function 154 inputs the correct answer data "lung cancer" and the doctor ID "A001" of the doctor who input the correct answer data, the image ID "XXX1", the modality ID "XXX", and the patient ID "0001". ” and stored in the useful data 143.

Next, the procedure of processing by the information collection system 1 according to the first embodiment will be explained using FIGS. 14, 15, and 16. FIG. 14 is a flowchart illustrating an example of the acquisition process according to the first embodiment. Here, each process from step S101 to S120 in FIG. 14 is realized, for example, by the processing circuit 15 reading out a program corresponding to the control function 151 from the storage circuit 14 and executing it. Note that the acquisition process shown in FIG. 14 is executed at any time, for example, when it is detected that the learning process or the discrimination process has been executed in the learning device 30.

For example, as shown in FIG. 14, in the information collection system 1 according to the present embodiment, the processing circuit 15 first obtains the calculation result of the intermediate layer from the learning device 30 through the network NW (step S101). Next, the processing circuit 15 stores the obtained calculation result in the calculation result information 141 (step S102).

Subsequently, the processing circuit 15 performs statistical processing on the calculation result stored in the calculation result information 141 (S103), and stores it in the statistical data 142 (S104).

Next, the processing circuit 15 determines whether processing for all filters has been completed (step S110). When the processing circuit 15 determines that the processing for all filters has not been completed (No in step S110), the processing circuit 15 moves to step S101 and repeats the processing.

On the other hand, if the processing circuit 15 determines that the processing for all filters has been completed (Yes at step S110), it determines whether the processing for all layers has been completed (step S120). If the processing circuit 15 determines that the processing for all layers has not been completed (No in step S120), the processing circuit 15 moves to step S101 and repeats the processing. On the other hand, if the processing circuit 15 determines that the processing for all layers has been completed (Yes at step S120), the processing circuit 15 ends the processing.

Next, a determination process based on newly acquired calculation results will be described. FIG. 15 is a flowchart illustrating an example of the determination process according to the first embodiment. Here, steps S201, S220, and S230 in FIG. 15 are realized, for example, by the processing circuit 15 reading out a program corresponding to the control function 151 from the storage circuit 14 and executing it. Further, steps S202, S210, and S211 are realized, for example, by the processing circuit 15 reading out a program corresponding to the determination function 152 from the storage circuit 14 and executing it. Note that the determination process shown in FIG. 15 is executed whenever the information collection process shown in FIG. 14 is executed, for example.

For example, as shown in FIG. 14, in the information collection system 1 according to the present embodiment, the processing circuit 15 first obtains the calculation result of the intermediate layer from the learning device 30 through the network NW (step S201).

Next, the processing circuit 25 compares the obtained calculation result with the statistical data 142 (S202), and determines whether the conditions for determining a rare case are satisfied, for example, whether the calculation result is equal to or higher than a threshold value. (Step S210). If the processing circuit 15 determines that the determination result is not equal to or greater than the threshold (No in step S210), the processing circuit 15 moves to step S220.

On the other hand, if the processing circuit 25 determines that the determination result is greater than or equal to the threshold (Yes in step S210), the processing circuit 25 stores, for example, medical information corresponding to the calculation result in the useful data 143 as the determination result (step S211). After that, the process moves to step S220.

Next, the processing circuit 15 determines whether processing for all filters has been completed (step S220). If the processing circuit 15 determines that the processing for all filters has not been completed (No in step S220), the processing circuit 15 moves to step S201 and repeats the processing.

On the other hand, if the processing circuit 15 determines that the processing for all filters has been completed (Yes at step S220), it determines whether the processing for all layers has been completed (step S230). When the processing circuit 15 determines that the processing for all layers has not been completed (No in step S230), the processing circuit 15 moves to step S201 and repeats the processing. On the other hand, if the processing circuit 15 determines that the processing for all layers has been completed (Yes at step S230), the processing circuit 15 ends the processing.

Next, a process for accepting input of correct data for useful data will be described. FIG. 16 is a flowchart illustrating an example of the reception process according to the first embodiment. Here, steps S301, S302, and S320 in FIG. 16 are realized, for example, by the processing circuit 15 reading out a program corresponding to the output function 153 from the storage circuit 14 and executing it. Further, steps S310 and S311 are realized, for example, by the processing circuit 15 reading out a program corresponding to the reception function 154 from the storage circuit 14 and executing it. Note that the reception process shown in FIG. 16 is executed, for example, in response to a trigger for starting the reception process. The trigger for starting the reception process may be, for example, an operation received from a CAD administrator via the input I/F 12, or may be received from the terminal 40 or the like via the communication I/F 11.

For example, as shown in FIG. 16, in the information collection system 1 according to this embodiment, the processing circuit 15 first acquires a record stored in the useful data 143 (step S301). Next, the processing circuit 15 generates a reception screen based on the acquired useful data and outputs it to the display 13 or the like of the information collection device 10 (S302). Then, the processing circuit 15 determines, for example, whether or not input of correct data has been received from the input I/F 12 (step S310). If the processing circuit 15 determines that input of correct data has not been accepted (No at step S310), it waits until input of correct data is accepted.

If the processing circuit 15 determines that input of correct data has been received (Yes at step S310), the processing circuit 15 stores the correct data in the useful data 143 (step S311). After that, the process moves to step S320.

Next, the processing circuit 15 determines whether processing for all useful data has been completed (step S320). When the processing circuit 15 determines that the processing for all useful data has not been completed (No in step S320), the processing circuit 15 moves to step S301 and repeats the processing. On the other hand, when the processing circuit 15 determines that processing for all useful data has been completed (Yes at step S320), the processing circuit 15 ends the processing.

As described above, the information collection system 1 according to the first embodiment can collect useful learning data. For example, the information collection system 1 according to the first embodiment can collect medical image data that corresponds to rare cases or data that a CAD administrator has determined to be useful for learning purposes.

(Second embodiment)
In the first embodiment, a configuration was described in which it is determined whether medical information satisfies the conditions based on the calculation results obtained from the intermediate layer of the trained model, but this is useful when preparing training data. The configuration for collecting such data is not limited to this. For example, if other medical image data is collected from a patient, etc. from whom data that falls under a rare case has been collected, even if the data itself does not fall under a rare case, the correct answer to the data that falls under a rare case May be useful in identifying data. Note that in this embodiment, "patient etc." is an example of a subject.

Therefore, in this embodiment, a configuration will be described in which useful information is collected by specifying a patient or the like whose medical image data has been collected. FIG. 17 is a diagram illustrating an example of functional blocks of the information collection system according to the second embodiment. In the following embodiments, the same parts as those shown in the drawings described above are denoted by the same reference numerals, and redundant explanation will be omitted.

As shown in FIG. 17, the information collection system 1a includes an information collection device 10a, a group of medical image diagnostic devices 20, an image storage device 21, a learning device 30, and a terminal 40. The information gathering device 10a includes a communication I/F 11, an input I/F 12, a display 13, a storage circuit 14a, and a processing circuit 15a.

In this embodiment, the storage circuit 14a stores calculation result information 141, statistical data 142, useful data 143a, and model data 144, as shown in FIG. The useful data 143a in this embodiment includes information on medical image data that is "suspected to be a rare case" and medical image data that "falls within the target range" as shown in FIG. Information regarding medical image data having the same "patient ID" is stored.

FIG. 18 is a diagram illustrating an example of useful data according to the second embodiment. As shown in FIG. 18, the useful data 143a stores the patient ID "0001" of the first record and a fourth record having the same patient ID. In this case, the useful data 143a stores that the corresponding reason for the fourth record is "patient match."

Returning to FIG. 17, the processing circuit 15 executes a control function 151, a determination function 152, an output function 153, a reception function 154, and a collection function 155. Here, the collection function 155 is an example of a collection unit.

A collection function 155 in the processing circuit 15 collects medical image data. For example, the collection function 155 acquires medical image data from the medical image diagnostic device group 20 or the image storage device 21 . Furthermore, the collection function 155 collects medical image data input from the learning device 30 into the input layer of the learned model 341 , 342 , or 343 . At this time, the collection function 155 acquires, for example, the image ID of the medical image data, the patient ID of the patient from whom the medical image data was collected, and the modality ID of the modality that collected the medical image data. Additionally, the collection function 155 may further acquire the model ID of the trained model into which the medical image data has been input.

The collection function 155 also determines, for example, whether a patient ID that matches the patient ID of the collected medical image data is stored in the useful data 143a. When the collection function 155 determines that a patient ID that matches the patient ID corresponding to the collected medical image data is stored in the useful data 143a, the collection function 155 stores information regarding the medical image data corresponding to the patient ID as the useful data. 143a.

For example, when the collection function 155 collects medical image data with a patient ID "0001", a modality ID "YYY", and a model ID "M003", the matching patient ID "0001" is the first item in the useful data 143a shown in FIG. It is determined that the data is stored in the first record. In this case, the collection function 155 stores information regarding the collected medical image data in the fourth record of the useful data 143a, as shown in FIG.

In this embodiment, the medical image collection process by the collection function 155 is executed whenever the control function 151 executes the information collection process, for example.

As described above, the information collection system 1 according to the second embodiment collects useful data such as medical image data that corresponds to rare cases and data that the CAD administrator has determined to be useful for learning purposes. Medical image data collected from the same subject as the data can be collected. This makes it possible to support the identification of cases in medical image data that corresponds to rare cases, data that the CAD administrator has determined to be useful for learning, and the like.

Note that the collection function 155 in this embodiment may, for example, anonymize the data corresponding to the patient ID when storing information regarding medical image data matching the patient ID in the useful data 143a. For example, when registering information regarding medical image data in the useful data 143a, the collection function 155 may replace the patient ID with a random character string so that the patient cannot be identified. In addition, the collection function 155 may, for example, replace the "k" patient IDs corresponding to other medical image data matching the case etc. with the same character string in addition to the patient ID corresponding to the medical image data concerned. "k-anonymization" may be realized by this.

(Third embodiment)
In the first embodiment described above, a configuration has been described in which the determination function 152 determines whether or not to collect calculation results using the statistical data 142. On the other hand, in the third embodiment, a case will be described in which the determination function determines a calculation result using a trained model.

FIG. 19 is a diagram illustrating an example of functional blocks of the information collection system according to the third embodiment. As shown in FIG. 19, the information collection system 1b includes an information collection device 10b, a group of medical image diagnostic devices 20, an image storage device 21, a learning device 30, and a terminal 40. The information gathering device 10b includes a communication I/F 11, an input I/F 12, a display 13, a storage circuit 14b, and a processing circuit 15b.

For example, as shown in FIG. 19, the storage circuit 14b further stores a learned model 145 in addition to calculation result information 141, statistical data 142, useful data 143, and model data 144. The trained model 145 is a trained model that is incorporated into a determination process executed by the processing circuit 15b that outputs whether or not medical image data corresponding to the calculation result is useful based on the input calculation result. be. The trained model 145 is configured by a multilayer neural network including an input layer, a plurality of intermediate layers, and an output layer, similar to the trained models 341, 342, and 343 shown in FIG. 5, for example. The trained model 341 is generated, for example, by the model generation function 156 described later.

Further, the processing circuit 15b executes a control function 151, a determination function 152b, an output function 153, a reception function 154, and a model generation function 156. The determination function 152b executes a determination process on the input calculation result based on the trained model 145. Specifically, the determination function 152b inputs the calculation result to the learned model 145, thereby executing a determination process to determine whether or not the medical image data corresponding to the calculation result is useful. Then, the determination function 152b outputs the determination result output from the learned model 145 to the output function 153.

The model generation function 156 executes processing related to generation of the trained model 145. For example, the model generation function 156 first receives input from the control function 151 of learning data including a set of a calculation result, which is input data, and information indicating whether or not the data is useful, which is correct data. The model generation function 156 causes the model to learn based on the learning data so as to determine correct data such as information indicating whether it is useful or not from the calculation results included in the learning data.

An example of the configuration of the information collection device 10b according to the third embodiment has been described above. With this configuration, the information collecting device 10b can determine whether medical image data is useful data based on the learned model based on the calculation result.

(Fourth embodiment)
In the first embodiment described above, a configuration is described in which the average value μ and standard deviation σ of the calculation results are used as statistical data. In contrast, in the fourth embodiment, a case will be described in which the relationship between the first calculation result and the second calculation result is used as the statistical data. FIG. 20 is a diagram illustrating an example of distribution of calculation results according to the fourth embodiment.

FIG. 20 shows the correlation distribution between calculation result 1, which is the first value, and calculation result 2, which is the second value. As shown in FIG. 20, most of the pairs of calculation results 1 and 2 are distributed along the range 98. In this case, the set 99a of calculation results 1 and 2 included in the range 98 is not determined to be a calculation result based on a rare case. On the other hand, pairs 99b and 99c of calculation results 1 and 2 that deviate from the range 98 are determined to be calculation results based on rare cases.

As described above, in the configuration according to the fourth embodiment, the information gathering device can determine the validity of the determination result using the learned model based on the relationship between the plurality of calculation results. Although FIG. 20 describes a configuration in which it is determined whether or not it is a rare case based on the relationship between two calculation results, the present invention is not limited to this, and the determination is made based on the relationship between three or more calculation results. It may be configured like this.

According to at least one embodiment described above, useful learning data can be collected.

Although several embodiments of the invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included within the scope and gist of the invention as well as within the scope of the invention described in the claims and its equivalents.

For example, although a configuration has been described in which medical image data is collected as medical information, the present invention is not limited to this, and may be a configuration that collects medical information other than images such as an electrocardiogram or audio data. Furthermore, although a configuration in which the learning device 30 has a plurality of trained models 341, 342, and 343 has been described, a configuration in which the learning device 30 has only a single trained model 341 may be used.

Further, the reception function 154 may further receive input from the user of a condition for determining whether or not the data is useful data, such as a condition such as "from -σ to the average value μ". In this case, the determination function 152 may determine the calculation result using the input conditions.

Further, the reception function 154 may further store in the useful data 143, for example, a doctor ID that identifies the doctor who has registered the correct answer data. For example, if multiple doctors input different cases as correct answer data for the same medical image data, the reception function 154 stores the case input by the largest number of doctors in the useful data 143 as correct answer data. It's okay. Further, the reception function 154 may determine the doctor's evaluation in advance, and store the case input by the doctor with the highest evaluation in the useful data 143 as correct data.

The term "processor" used in the description of each embodiment above refers to, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an Application Specific Integrated Circuit (ASIC), a programmable Refers to circuits such as logic devices (e.g., Simple Programmable Logic Device (SPLD), Complex Programmable Logic Device (CPLD), and Field Programmable Gate Array (FPGA)). do. Here, instead of storing the program in the memory, the program may be directly incorporated into the circuit of the processor. In this case, the processor realizes its functions by reading and executing a program built into the circuit. Furthermore, each processor of this embodiment is not limited to being configured as a single circuit for each processor, but may also be configured as a single processor by combining multiple independent circuits to realize its functions. good.

Here, the program to be executed by the processor is provided by being pre-installed in a ROM (Read Only Memory), a storage unit, or the like. This program is a file in a format that can be installed or executable on these devices, such as CD (Compact Disk)-ROM, FD (Flexible Disk), CD-R (Recordable), DVD (Digital Versatile Disk), etc. It may be provided recorded on a computer readable storage medium. Further, this program may be stored on a computer connected to a network such as the Internet, and may be provided or distributed by being downloaded via the network. For example, this program is composed of modules including each functional section. In actual hardware, a CPU reads a program from a storage medium such as a ROM and executes it, so that each module is loaded onto the main storage device and generated on the main storage device.

1 Information collection system 10 Information collection device 20 Medical image diagnosis device group 21 Image storage device 30 Learning device 40 Terminal

Claims (11)

In a first trained model that includes one or more intermediate layers that perform calculation processing on input information and an output layer that outputs diagnostic results based on the calculation results of the intermediate layers, the intermediate layer outputs a first storage unit that stores information regarding calculation results;
an acquisition unit that acquires a calculation result output by the intermediate layer in response to medical information being input to the input layer of the first trained model;
Compare the relationship between the calculation result acquired by the acquisition unit and a range determined from information regarding the calculation result stored in the first storage unit, and determine whether the calculation result acquired by the acquisition unit deviates from the range. a determination unit that determines that the medical information corresponds to a rare case if the
A second storage unit that stores information regarding the medical information corresponding to a calculation result in which it is determined that the medical information corresponds to a rare case. The medical information collection system according to claim 1, wherein the acquisition unit stores information including the acquired calculation result in the first storage unit as information regarding the calculation result. an output unit that outputs information regarding the medical information stored in the second storage unit;
Claim 2 further comprising: a reception unit that receives from a user an input of a judgment result regarding the medical information outputted by the output unit, and stores the accepted judgment result in the second storage unit in association with the medical information. The medical information collection system described in . 4. The medical information collection system according to claim 3, wherein the determination unit stores the medical information and the determination result in the second storage unit in association with subject information regarding the subject from whom the medical information has been collected. The acquisition unit further acquires subject information regarding the subject of the medical information input to the input layer, and subject information that matches the acquired subject information is stored in the second storage unit. 5. The medical information collection system according to claim 4, wherein, in the case where the medical information is inputted to the input layer in association with the matching subject information, the second storage unit stores information regarding the medical information input to the input layer in association with the matching subject information. The medical information collection system according to claim 4 or 5, wherein the reception unit stores the medical information related to the subject, the medical information, and the judgment result in anonymized form. The medical information collection system according to any one of claims 3 to 6, wherein the reception unit stores the medical information and the judgment result in the second storage unit in association with the user. The acquisition unit acquires the calculation results output by the intermediate layer from the plurality of first trained models,
Any one of claims 3 to 7, wherein the second storage unit stores the calculation results acquired by the acquisition unit in association with information regarding the plurality of first trained models that output the calculation results. The medical information collection system described in . further comprising a collection unit that collects the medical information from a plurality of medical image diagnostic devices,
The reception unit stores the medical information and the judgment result in association with information regarding the medical image diagnostic apparatus from which the collection unit has collected the medical information. Medical information collection system. The medical information collection system according to any one of claims 2 to 9, wherein the acquisition unit stores statistical data obtained by statistically processing the acquired calculation results in the first storage unit as information regarding the calculation results. . In a first trained model that includes one or more intermediate layers that perform calculation processing on input information and an output layer that outputs diagnostic results based on the calculation results of the intermediate layers, the intermediate layer outputs a first storage unit that stores information regarding calculation results;
an acquisition unit that acquires a calculation result output by the intermediate layer in response to medical information being input to the input layer of the first trained model;
Compare the relationship between the calculation result acquired by the acquisition unit and a range determined from information regarding the calculation result stored in the first storage unit, and determine whether the calculation result acquired by the acquisition unit deviates from the range. a determination unit that determines that the medical information corresponds to a rare case if the
a second storage unit that stores information regarding the medical information corresponding to a calculation result in which it is determined that the medical information corresponds to a rare case.

Images