Analysis device, analysis method, and medical image processing device

Abstract

To improve the detection relevance ratio and medical treatment efficiency of metastatic cancer.SOLUTION: An analysis device includes: a target organ analysis unit; and a related organ specification unit. The target organ analysis unit performs analysis on the basis of medical images of a target organ and acquires related organ information including at least the names of candidate related organs being candidates of the related organ related to the target organ and reliability of the candidate related organs. The related organ specification unit specifies the related organ related to the target organ from the reliability of the candidate related organs on the basis of the related organ information. Consequently, the analysis device improves the detection relevance ratio and medical treatment efficiency of metastatic cancer.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The embodiments disclosed in this specification and the drawings relate to an analysis device, an analysis method, and a medical image processing device. [Background technology]

[0002] In medical clinical research and applications, performing analysis based on medical images collected by medical image acquisition devices can assist doctors in quickly locating organs, tissues, tumors, etc.

[0003] In addition, in medical clinical research and applications, a primary tumor in one organ may metastasize to other organs, which is called metastatic cancer. Detection and segmentation of metastasis of malignant tumors is an important auxiliary function for doctors in diagnosis, staging, and treatment evaluation, and is an important task in medical image processing.

[0004] In the conventional technology, when a subject may have a metastatic lesion, a doctor must check the medical images of the relevant organs sequentially based on experience, manually search for and identify possible metastatic lesions, and perform segmentation and measurement. In this process, first, medical examinations are sequentially performed on the subject's relevant organs using medical image acquisition devices such as ultrasound diagnostic devices, X-ray CT (Computed Tomography) devices, and MRI (Magnetic Resonance Imaging) devices to collect medical images such as ultrasound images, CT images, and MR images. After that, the collected medical images must be interpreted one by one by a doctor, which is relatively time-consuming for both the subject and the doctor. In addition, when a doctor manually interprets images, there is a possibility that the doctor may overlook a metastatic lesion, which may affect the diagnosis.

[0005] Therefore, in the prior art, fully automatic lesion detection and segmentation algorithms for each organ have been developed. However, existing fully automatic lesion detection and segmentation algorithms have a certain degree of false positive results. In addition, when the corresponding fully automatic lesion detection and segmentation algorithms are performed for all organs (including organs without lesions), the number of false positives increases due to the accumulation of errors. This requires doctors to manually recheck these false positive lesions, which brings extra work to doctors.

[0006] Therefore, for the detection and segmentation of metastatic cancer, there is a need for technology that can analyze, detect, and segment metastatic cancer quickly and with a high accuracy rate. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] U.S. Patent No. 8,855,387 Summary of the Invention [Problem to be solved by the invention]

[0008] One of the problems to be solved by the embodiments disclosed in this specification and the drawings is to improve the detection accuracy rate and diagnostic efficiency of metastatic cancer. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problem. Problems corresponding to the effects of each configuration shown in the embodiments described later can also be positioned as other problems. [Means for solving the problem]

[0009] The analysis device according to the present embodiment includes a target organ analysis unit and an associated organ identification unit. The target organ analysis unit performs an analysis based on a medical image of the target organ, and obtains associated organ information including at least the names of candidate associated organs that are candidates for associated organs related to the target organ, and the reliability of the candidate associated organs. The associated organ identification unit identifies associated organs related to the target organ from the reliability of the candidate associated organs based on the associated organ information. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a block diagram showing the configuration of an analysis device according to the first embodiment. [Diagram 2] FIG. 2 is a flowchart showing the process of the analysis device according to the first embodiment. [Diagram 3] FIG. 3 is a diagram for explaining a specific example of the analysis device according to the first embodiment. [Figure 4] FIG. 4 is a block diagram showing the configuration of an analysis device according to the second embodiment. [Diagram 5] FIG. 5 is a flowchart showing the process of the analysis device according to the second embodiment. [Figure 6] FIG. 6 is a block diagram showing the configuration of an analysis device according to the third embodiment. [Figure 7] FIG. 7 is a flowchart showing the process of the target organ analysis unit of the analysis apparatus according to the third embodiment. [Figure 8A] FIG. 8A is a diagram for explaining a specific example of the analysis device according to the third embodiment. [Figure 8B] FIG. 8B is a diagram for explaining a specific example of the analysis device according to the third embodiment. [Figure 9A] FIG. 9A is a block diagram showing a configuration of an analysis device according to a modified example of the third embodiment. [Figure 9B] FIG. 9B is a flowchart showing the processing of the target organ analysis unit of the analysis device according to the modified example of the third embodiment. [Figure 9C]FIG. 9C is a diagram for explaining presented information of an analysis device according to a modified example of the third embodiment. [Figure 10] FIG. 10 is a block diagram showing the configuration of an analysis device according to the fourth embodiment. [Figure 11] FIG. 11 is a flowchart showing the process of the analysis device according to the fourth embodiment. [Figure 12] FIG. 12 is a diagram for explaining an example of the analysis device according to the fourth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, embodiments of an analysis device, an analysis method, and a medical image processing device will be described with reference to the drawings.

[0012] (First embodiment) [Configuration of analysis device 100] First, the configuration of an analysis device 100 according to the first embodiment will be described with reference to Fig. 1. Although the analysis device includes various components, Fig. 1 shows only the components related to the technical concept of the present embodiment, and the other components are omitted.

[0013] As shown in FIG. 1, the analysis device 100 according to the first embodiment includes a target organ analysis unit 101 and an associated organ identification unit 102. For example, the analysis device 100 is a computer, and includes a processing circuit for executing each processing function, a storage circuit for storing each processing function in the form of a program executable by a computer, an input interface, and a display. For example, the processing circuit of the analysis device 100 includes a target organ analysis unit 101 and an associated organ identification unit 102. Each processing function executed by the components of the processing circuit is recorded in the storage circuit, for example, in the form of a program executable by a computer. The processing circuit is a processor that reads each program from the storage circuit and executes it to realize the processing function corresponding to each program.

[0014] The term "processor" used in the above description means a circuit such as a central processing unit (CPU), a micro controller unit (MCU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)). When the processor is a CPU, for example, the processor realizes a function by reading and executing a program stored in a memory circuit. On the other hand, when the processor is an ASIC, for example, instead of storing a program in a memory circuit, a program is directly incorporated in the circuit of the processor. Note that each processor in this embodiment is not limited to being configured as a single circuit for each processor, but may be configured as one processor by combining multiple independent circuits to realize its function. Furthermore, multiple components may be integrated into one processor to realize its function.

[0015] The target organ analysis unit 101 performs an analysis based on, for example, a medical image of the target organ, and acquires related organ information including at least the name of a candidate related organ that is a candidate for a related organ related to the target organ, and the reliability of the candidate related organ.

[0016] Here, the target organ is an organ of a subject, and is an initial target organ for which the analysis device 100 performs an analysis of metastatic cancer, and may be, for example, the colon, liver, lungs, lymph nodes, etc. The target organ may also be an organ known to have a cancer focus, and an associated organ related to the target organ refers to an organ in which a cancer focus may exist and in which the cancer focus is associated with the cancer focus of the target organ, i.e., a metastatic focus. The initial target organ may be specified by a user (e.g., a doctor), or may be specified by performing known image processing, such as image segmentation, on a medical image including the target organ.

[0017] The reliability of a candidate related organ included in the related organ information is a quantified value indicating the possibility that the candidate related organ is related to the target organ.

[0018] Moreover, the medical image of the target organ may be a medical image including the entire target organ. The medical image of the target organ may be a medical image such as an ultrasound image, a CT image, or an MR image captured by a medical image collecting device such as an ultrasound diagnostic device, an X-ray CT (Computed Tomography) device, or an MRI (Magnetic Resonance Imaging) device. That is, the medical image may be a three-dimensional image or a two-dimensional image. The medical image of the target organ may be acquired from a medical image collecting device by a communication method, for example, or may be acquired by reading a storage device that previously stores the medical image of the target organ, or may be acquired by reading a medical image processed by another component. Moreover, in this embodiment, the medical image of the target organ can be acquired at any time as necessary, and the analysis device 100 can acquire any necessary medical image of the target organ.

[0019] Here, as a method for the target organ analysis unit 101 to acquire related organ information, for example, the target organ analysis unit 101 acquires the related organ information by analyzing based on the examination images of the subject and the corresponding diagnostic information of the doctor, and multiple follow-up examination images of the subject and the corresponding diagnostic information of the doctor.

[0020] As an example, when a cancer lesion is found in the liver as a target organ in a test image of a subject, and in addition, a cancer lesion is also found to be present in the lungs from the test image of the subject and / or the corresponding diagnostic information of a doctor, the target organ analysis unit 101 can determine the lungs as the target organ, i.e., an associated organ of the liver.

[0021] As another example, in a case where a test image of a subject shows a cancer focus in the liver as a target organ, and also a cancer focus is present in the lungs, and a test image of a follow-up test of the subject shows a cancer focus in the lymph nodes, the target organ analysis unit 101 can acquire analysis results such as "the candidate associated organs of the target organ "liver" are the lungs and lymph nodes," "the candidate associated organs of the target organ "liver" are the lungs," "the candidate associated organs of the target organ "liver" are the lymph nodes," "the candidate associated organs of the target organ "liver" are the lungs and lymph nodes, and the candidate associated organs of the lungs are the lymph nodes" as the related organ information.

[0022] Similarly, the reliability of the candidate related organ included in the related organ information is obtained by various known methods based on the examination image of the subject and the corresponding doctor's diagnostic information, and the subject's multiple follow-up examination images and the corresponding doctor's diagnostic information. In this way, the target organ analysis unit 101 can perform an analysis based on the medical image of the target organ and obtain related organ information including at least the name of the candidate related organ that is a candidate of the related organ related to the target organ and the reliability of the candidate related organ.

[0023] In addition, the method for the target organ analysis unit 101 to acquire the relevant organ information may be to perform deep learning based on the examination image of the subject and the corresponding doctor's diagnosis information, and multiple follow-up examination images of the subject and the corresponding doctor's diagnosis information, and then acquire the relevant organ information based on a deep learning model by deep learning. Note that the method for acquiring the relevant organ information based on deep learning will be described in detail in the third embodiment.

[0024] In addition, the representation format of the related organ information may be an image superimposed on a medical image of the target organ, or a table or database that associates at least the target organ, the names of the candidate related organs, and the reliability of the candidate related organs, or further, text data containing information such as at least the target organ, the names of the candidate related organs, and the reliability of the candidate related organs.

[0025] Furthermore, the associated organ identifying unit 102 identifies an associated organ associated with the target organ from the reliability of the candidate associated organ based on the associated organ information, for example.

[0026] As a method for the associated organs specified by the associated organ specifying unit 102, for example, when all the associated organs whose names are included in the associated organ information or the associated organs whose reliability included in the associated organ information is equal to or greater than a predetermined value are set as candidate associated organs, the unique value or the maximum value of the reliability of each candidate associated organ from the reliability of the associated organs included in the associated organ information is set as the reliability of the candidate associated organ, and the candidate associated organ with the highest reliability or the reliability equal to or greater than a predetermined threshold is specified as the associated organ. Also, when a lesion has already been detected in a certain candidate associated organ, the reliability of this candidate associated organ is set to 0.

[0027] The number of associated organs identified by the associated organ identifying unit 102 may be one, a predetermined number, or a plurality of organs whose reliability is equal to or greater than a predetermined threshold. The associated organ identifying unit 102 can identify that there is no associated organ when the reliability included in the associated organ information is all less than a predetermined lower limit.

[0028] In addition, the above-mentioned specified values, specified threshold values, and specified lower limits can be set or changed as appropriate using a number of methods, and can, for example, be set by user input (doctor's input) or automatically set based on medical images, organs, user preferences, reliability calculation methods, etc.

[0029] [Processing by analysis device 100] The following describes an analysis method applied by the analysis device 100 according to the first embodiment. Fig. 2 is a flowchart showing the processing of the analysis device 100 according to the first embodiment. Fig. 3 is a diagram for explaining a specific example of the analysis device 100 according to the first embodiment.

[0030] In the analysis device 100 according to the first embodiment, after starting the process, in step S100 in FIG. 2, the target organ analysis unit 101 performs an analysis based on a medical image of the target organ, and acquires associated organ information including at least the name of a candidate associated organ that is a candidate of an associated organ related to the target organ and the reliability of the candidate associated organ. Then, the process of the analysis device 100 proceeds to step S200. For example, in step S100, the target organ analysis unit 101 acquires associated organ information such as "associated organ: liver (0.85)" and "associated organ: liver (0.78), lung (0.66)" as shown in the lower part of FIG. 3 based on an abdominal medical image including a colon as a target organ as shown in the upper part of FIG. 3. Here, the numerical value in parentheses shown in the lower part of FIG. 3 indicates the reliability of the candidate associated organ.

[0031] In step S200 of FIG. 2, the associated organ identifying unit 102 identifies an associated organ related to the target organ from the reliability of the candidate associated organ based on the associated organ information acquired by the target organ analysis unit 101 in step S100. At this time, for example, the associated organ identifying unit 102 presents the identified associated organ to the user (doctor) by displaying it on a display. Here, in the example shown in FIG. 3, the "liver" and "lung" can be identified as candidate associated organs based on the associated organ information acquired by the target organ analysis unit 101. For example, in the lower part of FIG. 3, it is shown that the reliability of the candidate associated organ "liver" has two values, 0.85 and 0.78, and the reliability of the candidate associated organ "lung" is a unique value, that is, 0.66. Therefore, when a condition is adopted that the candidate associated organ with the highest reliability is identified as the associated organ, the associated organ identifying unit 102 can identify the candidate associated organ "liver" with a reliability of 0.85 as the associated organ. In addition, when the condition that the reliability of the candidate associated organ is equal to or greater than a predetermined threshold (e.g., 0.6) is met, the associated organ identification unit 102 can identify two candidate associated organs, the liver and the lungs, as the associated organs.

[0032] The analysis device 100 according to the first embodiment acquires related organ information including the reliability of a candidate related organ, which is a candidate of a related organ related to a target organ, based on a medical image of the target organ, and automatically detects and identifies related organs related to the target organ from the reliability of the candidate related organ. As a result, the analysis device 100 according to the first embodiment can improve the detection accuracy of metastatic cancer. Furthermore, compared to the conventional technology, the analysis device 100 according to the first embodiment does not need to execute a corresponding fully automatic lesion detection and segmentation algorithm for all organs (including organs without lesions), so that it is possible to avoid unnecessary and unnecessary calculations, avoid the accumulation of errors, and reduce false positive results. Furthermore, the analysis device 100 according to the first embodiment can reduce the process of doctors manually reconfirming these false positives, thereby reducing the workload of the doctors. Furthermore, the analysis device 100 according to the first embodiment does not need doctors to manually confirm related organs as in the conventional technology, and it is possible to avoid the burden on the doctors. Therefore, the analysis device 100 according to the first embodiment can improve the efficiency of medical treatment.

[0033] Second embodiment [Configuration of analysis device 100A] The configuration of analysis device 100A according to the second embodiment will be described with reference to Fig. 4. Fig. 4 is a block diagram showing the configuration of analysis device 100A according to the second embodiment.

[0034] 4, the analysis device 100A according to the second embodiment further includes a control unit 103 in addition to a target organ analysis unit 101 and an associated organ identification unit 102. Specifically, the processing circuit of the analysis device 100A includes the target organ analysis unit 101, the associated organ identification unit 102, and the control unit 103.

[0035] Here, the target organ analysis unit 101 and the associated organ identification unit 102 are the same as the target organ analysis unit 101 and the associated organ identification unit 102 in the first embodiment, respectively. Therefore, in this embodiment, a duplicated description will be omitted.

[0036] The control unit 103 of the analysis device 100A according to the second embodiment controls the process of repeatedly analyzing the relevant organ identified by the relevant organ identification unit 102 as a new target organ in the target organ analysis unit 101, and then identifying the next new relevant organ by the relevant organ identification unit 102, until the relevant organ identification unit 102 identifies that there is no relevant organ.

[0037] Furthermore, control unit 103 may, as a control unit for the entire analysis device 100A, be connected to each of the components of analysis device 100A and coordinate and control the operation of each of the components.

[0038] [Processing by analysis device 100A] Next, an analysis method applied by analysis device 100A according to the second embodiment will be described below. Fig. 5 is a flowchart showing the process of analysis device 100A according to the second embodiment.

[0039] After starting the process, analysis device 100A executes the process of steps S100 to S200 in Fig. 5, and then the process of analysis device 100A proceeds to step S300. Here, the process of steps S100 to S200 is the same as the process of steps S100 to S200 in Fig. 2, and therefore a detailed description thereof will be omitted here.

[0040] 5, analysis device 100A determines whether or not associated organ has been identified by associated organ identifying unit 102 as not having an associated organ (step S300) through control unit 103. If control unit 103 determines that associated organ has been identified by associated organ identifying unit 102 as not having an associated organ (step S300: YES), processing of analysis device 100A ends. On the other hand, if control unit 103 determines that associated organ has been identified by associated organ identifying unit 102 as having an associated organ (step S300: NO), processing of analysis device 100A proceeds to step S301.

[0041] 5, the control unit 103 controls the relevant organ identified by the relevant organ identifying unit 102 to be a new target organ. After that, the processing of the analysis device 100A returns to step S100, and steps S100 to S200 are repeatedly executed; that is, after the target organ analysis unit 101 analyzes this new target organ, the relevant organ identifying unit 102 identifies the next new relevant organ.

[0042] The analysis device 100A according to the second embodiment realizes the effects of the first embodiment. In addition, the analysis device 100A according to the second embodiment includes a control unit 103, and controls the process of "analyzing the relevant organ identified by the relevant organ identification unit 102 as a new target organ in the target organ analysis unit 101, and then identifying the next new relevant organ by the relevant organ identification unit" repeatedly until the relevant organ identification unit 102 identifies that there is no relevant organ. As a result, the analysis device 100A according to the second embodiment can automatically acquire all relevant organs based on the medical image of the target organ without requiring any operation by the user (doctor). Therefore, the analysis device 100A according to the second embodiment further improves the efficiency of medical care and the detection accuracy rate of metastatic cancer.

[0043] (Third embodiment) [Configuration of analysis device 100B] First, the configuration of analysis device 100B according to the third embodiment will be described with reference to Fig. 6. Fig. 6 is a block diagram showing the configuration of analysis device 100B according to the third embodiment.

[0044] As shown in Fig. 6, the analysis device 100B according to the third embodiment includes a target organ analysis unit 101, an associated organ identification unit 102, and a control unit 103, similar to the first embodiment. Here, the analysis device 100B according to the third embodiment is different from the first embodiment in that the target organ analysis unit 101 includes a target organ image range determination unit 1011 and a target organ lesion detection unit 1012. Note that the associated organ identification unit 102 is the same as the associated organ identification unit 102 in the first embodiment, and the control unit 103 is the same as the control unit 103 in the second embodiment, so that duplicated explanations will be omitted in this embodiment.

[0045] In this embodiment, the target organ image area determining unit 1011 determines the target organ image area for a medical image of the target organ, for example, automatically or manually by the user.

[0046] Here, the "target organ image range" is, for example, an image range that includes the entire target organ.

[0047] As a method for determining the target organ image range, for example, a user may specify (for example, perform a selection operation) the position of a certain part (for example, a lesion) in the target organ on the medical image of the subject, and then automatically determine the target organ image range using a known organ segmentation algorithm or organ detection algorithm. Alternatively, a user may specify a target organ on the medical image of the subject, and then automatically determine the target organ image range using a known organ segmentation algorithm or organ detection algorithm. Furthermore, a user may specify an image range including the target organ on the medical image of the subject, and then optimize (for example, enlarge, translate, etc.) the image range including the target organ using a known organ segmentation algorithm or organ detection algorithm to automatically determine the target organ image range region including the complete target organ.

[0048] In addition, the target organ lesion detection unit 1012, for example, detects all lesions of the target organ within the target organ image range determined by the target organ image range determination unit 1011, and acquires lesion position information for each detected lesion and associated organ information corresponding to each lesion.

[0049] Here, the lesion position information indicates the relationship between the lesion and the medical image (or the target organ image range), and other similar information, such as mask and radiomics information, can be used.

[0050] Also in the third embodiment, the relevant organ related to the target organ refers to an organ in which a cancer focus may exist and the cancer focus is a focus related to the cancer focus of the target organ, i.e., a metastatic focus, as described above. Also in the third embodiment, the relevant organ information here includes at least the relevant organ name and reliability, as described above. Also in the third embodiment, the relevant organ information corresponds to the target organ as described above, and further corresponds to the focus of the target organ. In other words, when multiple focus exists in the same organ, the multiple focus may be related to the same organ or different other organs.

[0051] Furthermore, the method in which the target organ lesion detection unit 1012 detects all lesions in the target organ based on the target organ image range and acquires lesion position information for each lesion and associated organ information corresponding to each lesion may be, for example, a method based on deep learning. That is, the target organ lesion detection unit 1012 detects all lesions in the target organ within the target organ image range determined by the target organ image range determination unit 1011 based on a deep learning model, and acquires lesion position information for each lesion and associated organ information corresponding to each lesion for all detected lesions. Here, the deep learning model is an example of a "first deep learning model".

[0052] For example, one deep learning model can be trained for each organ in advance or in real time. The deep learning model is trained with training data to detect all lesions in the target organ for each organ, and output lesion location information for each detected lesion and associated organ information corresponding to each lesion. Here, the training data includes, for example, an examination image of a subject and corresponding doctor's diagnosis information, and multiple follow-up examination images of the subject and corresponding doctor's diagnosis information. Here, the training data is an example of "first training data".

[0053] Also, a deep learning model can be trained for each organ individually, in advance or in real time, where the deep learning model is trained with the training data to detect only lesions for each organ.

[0054] In addition, when acquiring the lesion of an organ based on a deep learning model, the reliability can be defined as a value representing the possibility that the designated organ is an associated organ of the lesion, for example, for one lesion and a designated organ. For example, it is defined as shown in the following formula.

[0055]

number

[0056] Here, A i indicates the similarity between the lesion location information output by the deep learning model and the i-th training data used to train the deep learning model. The training data is annotated in advance and is also called GT (Ground Truth) or annotation GT. A i The value ranges from 0 to 1. i A value of 1 indicates that they are exactly the same, and A i A value of 0 indicates that they are completely different.

[0057] B i indicates whether the designated organ (related organ) output by the deep learning model matches the related organ in the training data mentioned above.i If the value of is 1, it indicates that the designated organ (related organ) output by the deep learning model matches the related organ in the training data, and B i If the value is 0, it indicates that the designated organ (associated organ) output by the deep learning model does not match the associated organ in the training data.

[0058] n is the total number of training data used in training the deep learning model.

[0059] It should be noted that the above definition is merely an example, and the definition of the reliability of the associated organ is not limited to this, and various existing definitions may be used.

[0060] [Processing of analysis device 100B] Next, an analysis method applied by the analysis device 100B according to the third embodiment will be described. Fig. 7 is a flowchart showing the processing of the target organ analysis unit 101 of the analysis device 100B according to the third embodiment. Figs. 8A and 8B are diagrams for explaining a specific example of the analysis device 100B according to the third embodiment.

[0061] The flowchart shown in Fig. 7 is a flowchart showing the processing of the target organ analysis unit 101. That is, the target organ analysis unit 101 of the analysis device 100B according to the third embodiment performs the processing of steps S101 to S102 in Fig. 7, instead of step S100 in the first embodiment (Fig. 2) and the second embodiment (Fig. 5).

[0062] 7, in the target organ analysis unit 101, the target organ image range determination unit 1011 first determines the target organ image range for the medical image of the target organ automatically or manually by the user. After that, the process of the analysis device 100B proceeds to step S102.

[0063] For example, in step S101, it is assumed that the user clicks on the position of the colon on the medical image of the target organ using the input interface (the position indicated by the dot 801 in the image on the left side of FIG. 8A). In this case, the target organ image range determination unit 1011 segments or detects the organ at the position indicated by the dot 801, i.e., the target organ, using a known organ segmentation algorithm or organ detection algorithm. Then, the target organ image range determination unit 1011 crops the target organ image range (image range 802 surrounded by a dashed line in the image on the left side of FIG. 8A), which is an image range including the segmented or detected target organ, from the medical image of the target organ using a known image cropping. That is, the target organ image range determination unit 1011 cuts out a partial image (the image on the right side of FIG. 8A) from the medical image of the target organ.

[0064] In step S102 in FIG. 7, the target organ lesion detection unit 1012 detects all lesions of the target organ within the target organ image range, and acquires lesion position information for each lesion and associated organ information corresponding to each lesion.

[0065] For example, in step S102, the target organ lesion detection unit 1012 inputs the target organ image range (the image range on the left side of FIG. 8B) to a pre-trained deep learning network (deep learning model) to obtain information as shown on the right side of FIG. 8B. Specifically, in the example shown on the right side of FIG. 8B, the target organ lesion detection unit 1012 detects all lesions "lesion 1" and "lesion 2" of the target organ within the target organ image range determined by the target organ image range determination unit 1011 based on the deep learning model. Here, for the detected lesion "lesion 1", the target organ lesion detection unit 1012 obtains lesion position information "position information: (100, 84, 411)-(123, 122, 434)" of the "lesion 1" and related organ information "related organ: liver (0.85)" corresponding to the "lesion 1". In addition, for the detected lesion "Lesion 2", the target organ lesion detection unit 1012 acquires lesion position information of "Lesion 2" (position information: (40, 45, 450)-(55, 72, 465)) and associated organ information corresponding to "Lesion 2" (associated organs: liver (0.78), lung (0.66)).

[0066] Thereafter, the process of analysis device 100B proceeds to step S200 in the first embodiment (FIG. 2) and the second embodiment (FIG. 5).

[0067] In this way, the analysis device 100B according to the third embodiment crops an image range including the target organ from a medical image of the target organ, which eliminates the need to execute existing fully automatic lesion detection and segmentation algorithms for all organs (including organs without lesions) compared to conventional techniques, thereby improving the accuracy of detection of metastatic cancer. Also, the analysis device 100B according to the third embodiment crops an image range including the target organ from a medical image of the target organ, which reduces the image size, thereby avoiding unnecessary calculations and improving the efficiency of analysis, detection, and medical treatment.

[0068] (Modification of the third embodiment) This embodiment is not limited to the above-mentioned configuration, and may have various modifications. For example, Fig. 9A is a block diagram showing the configuration of an analysis device 100BB according to a modification of the third embodiment.

[0069] 9A, the analysis device 100BB according to the modification of the third embodiment includes a target organ analysis unit 101, a related organ identification unit 102, and a control unit 103, similar to the third embodiment, and the target organ analysis unit 101 includes a target organ image range determination unit 1011 and a target organ lesion detection unit 1012. Here, the target organ image range determination unit 1011, the target organ lesion detection unit 1012, the related organ identification unit 102, and the control unit 103 of the target organ analysis unit 101 in the modification of the third embodiment are all the same as the target organ image range determination unit 1011, the target organ lesion detection unit 1012, the related organ identification unit 102, and the control unit 103 of the target organ analysis unit 101 in the above-mentioned third embodiment, so that a duplicated description will be omitted in this embodiment.

[0070] Moreover, the analysis device 100BB according to the modified example of the third embodiment differs from the third embodiment in that the target organ analysis section 101 further includes a focus analysis section 1013.

[0071] The lesion analysis section 1013 performs an analysis on all lesions based on the lesion position information acquired by the target organ lesion detection section 1012, for example, to obtain lesion analysis information for each lesion.

[0072] Here, the lesion analysis information includes, but is not limited to, lesion contour segmentation, measurement value calculation regarding the tumor's major axis, minor axis, volume, etc., histogram calculation, texture analysis, and the like.

[0073] The lesion analysis information is presented to the user together with lesion location information and / or associated organ information (described later in Figure 9C), which contributes to the user's judgment on the reliability of the associated organ information and is advantageous in increasing the accuracy of the user's judgment on cancer metastasis.

[0074] Although not shown, the analysis device 100BB may present various information such as related organ information, lesion location information, lesion analysis information, etc., to the user by displaying it on a display. As a method for presenting various messages to the user, they may be presented to the user by light or warning sound, or may be presented to the user by voice and / or a screen (display), as necessary.

[0075] Fig. 9B is a flowchart showing the processing of the target organ analysis unit 101 of the analysis device 100BB according to the modification of the third embodiment. The flowchart shown in Fig. 9B is a flowchart showing the processing of the target organ analysis unit 101. That is, the target organ analysis unit 101 of the analysis device 100BB according to the modification of the third embodiment performs the processing of steps S101 to S103 in Fig. 9B instead of step S100 in the first embodiment (Fig. 2) and the second embodiment (Fig. 5).

[0076] As shown in Fig. 9B, the target organ analysis unit 101 of the analysis device 100BB first executes the process of steps S101 to S102, and then the process of the analysis device 100BB proceeds to step S103. Here, the operations of steps S101 to S102 are the same as the operations of steps S101 to S102 in Fig. 7, so that the overlapping explanation will be omitted in this embodiment.

[0077] In step S103 in Fig. 9B, the lesion analysis unit 1013 performs analysis on all lesions based on the lesion position information acquired by the target organ lesion detection unit 1012 in step S102, and acquires lesion analysis information for each lesion. At this time, the lesion analysis unit 1013 presents the acquired lesion analysis information to a user (doctor) by displaying it on a display together with the lesion position information and the related organ information. Thereafter, the processing of the analysis device 100BB proceeds to step S200 in the first embodiment (Fig. 2) and the second embodiment (Fig. 5).

[0078] In the example shown in FIG. 9C, in step S103, a cancer focus is present in the target organ "colon", all the focus "focus 1" and "focus 2" in the target organ "colon" are detected, and focus position information for each focus, related organ information corresponding to each focus, and focus analysis information for each focus are acquired. The related organ information includes at least the names of candidate related organs that are candidates for related organs related to the target organ "colon", "liver", "lung", and "lymph node", and the reliability of the candidate related organs, "0.90", "0.80", and "0.70". The focus analysis information includes at least information on the size of the focus, such as the long diameter (LD; also called long axis), short axis (SA; also called short axis), and volume of the focus.

[0079] In this case, in FIG. 9C, the name of the target organ "colon" and the reliability "1.0" and the lesion location information and lesion analysis information of the lesions "lesion 1" and "lesion 2" of the target organ "colon" are presented. In addition, the name of the candidate related organ "liver" and the reliability "0.90" and the lesion location information and lesion analysis information of the candidate related organ "liver" corresponding to the lesions "lesion 1" and "lesion 2" of the target organ "colon" are presented. In addition, the name of the candidate related organ "lung" and the reliability "0.80" and the lesion location information and lesion analysis information of the candidate related organ "lung" corresponding to the lesions "lesion 1" and "lesion 2" of the target organ "colon" are presented. In addition, the name of the related organ "lymph node" and the reliability "0.70" and the lesion location information and lesion analysis information of the candidate related organ "lymph node" corresponding to the lesions "lesion 1" and "lesion 2" of the target organ "colon" are presented.

[0080] The analysis device 100B according to the third embodiment and the analysis device 100BB according to the modification thereof realize the technical effect of the first embodiment. In addition, the analysis device 100B according to the third embodiment and the analysis device 100BB according to the modification thereof include a target organ image range determination unit 1011 and a target organ lesion detection unit 1012, so that all lesions in the target organ can be detected, and for all detected lesions, lesion position information for each lesion and associated organ information corresponding to each lesion can be obtained. In addition, the analysis device 100BB according to the modification of the third embodiment further includes a lesion analysis unit 1013, so that lesion analysis information for each lesion can be obtained. In this way, the analysis device 100B according to the third embodiment and the analysis device 100BB according to the modification thereof can obtain more detailed information on organs and lesions, so that the user can quickly make a judgment and perform medical treatment on metastatic cancer, and the accuracy and reliability are further improved.

[0081] (Fourth embodiment) [Configuration of analysis device 100C] The configuration of an analysis device 100C according to the fourth embodiment will be described with reference to Fig. 10. Fig. 10 is a block diagram showing the configuration of an analysis device 100C according to the fourth embodiment.

[0082] 10, an analysis device 100C according to the fourth embodiment includes a target organ analysis unit 101, an associated organ identification unit 102, and a control unit 103, similar to the modification of the third embodiment. Here, the target organ analysis unit 101 includes a target organ image range determination unit 1011, a target organ lesion detection unit 1012, and a lesion analysis unit 1013. However, a difference from the modification of the third embodiment is that the associated organ identification unit 102 includes a candidate associated organ identification unit 1021, a reliability calculation unit 1022, and an identification unit 1023. In addition, the target organ image range determination unit 1011, target organ lesion detection unit 1012, and lesion analysis unit 1013 of the target organ analysis unit 101, the associated organ identification unit 102, and the control unit 103 are identical to the target organ image range determination unit 1011, target organ lesion detection unit 1012, and lesion analysis unit 1013 of the target organ analysis unit 101, the associated organ identification unit 102, and the control unit 103 in the modified example of the third embodiment, so duplicated explanations will be omitted in this embodiment.

[0083] The candidate associated organ identifying unit 1021 identifies all candidate associated organs based on the associated organ information obtained by the target organ analyzing unit 101, for example.

[0084] The reliability calculation unit 1022 calculates the reliability of each of all the candidate related organs identified by the candidate related organ identification unit 1021, for example.

[0085] The identifying unit 1023 identifies, for example, a candidate associated organ having the highest reliability calculated by the reliability calculation unit 1022 or a candidate associated organ having a reliability equal to or greater than a predetermined threshold, as an associated organ.

[0086] [Processing of analysis device 100C] Next, an analysis method applied by the analysis device 100C according to the fourth embodiment will be described below. Fig. 11 is a flowchart showing the process of the analysis device 100C according to the fourth embodiment.

[0087] After starting the process, analysis device 100C executes the process of steps S101 to S103 in FIG. 11, and then the process of analysis device 100C proceeds to the process of steps S201 to S203, and then to the process of steps S300 to S301. Here, the process of steps S101 to S103 in FIG. 11 is equivalent to the process of step S100 in FIG. 5, and the process of steps S201 to S203 in FIG. 11 is equivalent to the process of step S200 in FIG. 5. In addition, the process of steps S101 to S103 in FIG. 11 is the same as the process of steps S101 to S103 in FIG. 9B, and the process of steps S300 to S301 is the same as the process of steps S300 to S301 in FIG. 5. For this reason, the overlapping detailed description will be omitted here, and only the process of steps S201 to S203 in FIG. 11 will be described.

[0088] In step S201 of FIG. 11, the candidate associated organ identifying unit 1021 of the associated organ identifying unit 102 identifies all candidate associated organs based on the associated organ information obtained in step S102, and then the processing of the analysis device 100C proceeds to step S202.

[0089] In step S202 of FIG. 11, the reliability calculation unit 1022 of the associated organ identification unit 102 calculates the reliability of each candidate associated organ for all candidate associated organs identified by the candidate associated organ identification unit 1021 in step S201, and then the processing of the analysis device 100C proceeds to step S203.

[0090] In step S203 of FIG. 11, the identification unit 1023 of the associated organ identification unit 102 identifies as the associated organ the candidate associated organ with the highest reliability calculated by the reliability calculation unit 1022 in step S202 or the candidate associated organ with a reliability equal to or greater than a predetermined threshold, and then the processing of the analysis device 100C proceeds to step S300.

[0091] (Example) Below, examples of the fourth embodiment will be described in combination with specific examples.

[0092] FIG. 12 is a diagram for explaining a specific example of the operation of analysis device 100C according to the fourth embodiment.

[0093] 12 is a medical image of a subject. In this embodiment, analysis of metastatic cancer is performed on the medical image of the subject by the analysis device 100C according to the fourth embodiment.

[0094] First, the analysis device 100C starts the circulation process indicated by "1" as shown in Fig. 12. The analysis device 100C determines the target organ image range based on the medical image of the subject by the target organ image range determining unit 1011, and in this embodiment, the image range of the target organ determined by the user clicking on the colon part of the medical image of the subject is set to the target organ as "colon" as in Fig. 8A (step S101 in Fig. 11).

[0095] Next, the target organ lesion detection unit 1012 detects all lesions in the colon, which is the target organ, and obtains lesion position information for each lesion and associated organ information corresponding to each lesion (step S102 in FIG. 11). For example, one lesion in the colon (the "colon" part is shown by a dot in the lower part of FIG. 12) is detected, and associated organ information of "Lesion 1 (lesion position: ####); associated organs: liver (0.8), lung (0.6), stomach (0.1)" is obtained.

[0096] Next, the lesion analysis unit 1013 performs lesion analysis on all lesions in the "colon" to obtain lesion analysis information (step S103 in FIG. 11).

[0097] Next, the candidate associated organ identifying unit 1021 identifies all the candidate associated organs based on the associated organ information (step S201 in FIG. 11). In this embodiment, all the candidate associated organs are the "liver", the "lung", and the "stomach".

[0098] Next, the reliability calculation unit 1022 calculates the reliability of each of all the candidate related organs (step S202 in FIG. 11). In this embodiment, the candidate related organs are the "liver", the "lung", and the "stomach", and the reliability is "0.8", "0.6", and "0.1", respectively.

[0099] Next, the identification unit 1023 identifies the most reliable candidate related organ or a candidate related organ whose reliability is equal to or greater than a predetermined threshold as a related organ (step S203 in FIG. 11). In this embodiment, since the condition is that the candidate related organ whose reliability is the highest is identified as a related organ, the "liver" has the highest reliability of "0.8" among the candidate related organs and is therefore identified as a related organ.

[0100] Next, the control unit 103 judges whether or not no relevant organ has been identified (step S300 in FIG. 11). In this embodiment, since the "liver" is identified as the relevant organ of the target organ "colon", the result of the judgment in step S300 is NO.

[0101] Next, the control unit 103 specifies the relevant organ "liver" as a new target organ (step S300 in FIG. 11), and the processing of the analysis device 100C returns to step S101.

[0102] Up to this point, analysis device 100C has completed the circulation process indicated by "1" in FIG.

[0103] Next, analysis device 100C starts the process of the loop indicated by "2" as shown in Fig. 12. The processes of the loops indicated by "2", "3", and "4" in Fig. 12 are the same as the specific process of the loop indicated by "1" described above. Therefore, only a brief explanation will be given below.

[0104] In the loop represented by "2" in Figure 12, the following processing is carried out:

[0105] First, the analysis device 100C determines the target organ image range of the "liver" which is a new target organ, based on a medical image of a subject, by the target organ image range determining unit 1011 (step S101 in FIG. 11).

[0106] Next, the target organ lesion detection unit 1012 detected two lesions in the target organ "liver" (the "liver" part is indicated by a dot in the center of Figure 12), and obtained the associated organ information "Lesion 1 (lesion location: ######); associated organs: liver (0.8), lung (0.5)" and "Lesion 2 (lesion location: %%%%%); associated organs: lymph nodes (0.4), lung (0.6), stomach (0.1)" (step S102 in Figure 11).

[0107] Next, the lesion analysis unit 1013 performs lesion analysis on all lesions in the "liver" to obtain lesion analysis information (step S103 in FIG. 11).

[0108] Next, the candidate associated organ identifying unit 1021 identifies all the candidate associated organs as the "liver", "lung", "lymph node", and "stomach" (step S201 in FIG. 11).

[0109] Next, the reliability calculation unit 1022 calculates the reliability of each of the candidate related organs, i.e., the "liver," "lungs," "lymph nodes," and "stomach," for all the candidate related organs. Among the candidate related organs, the "liver" has the highest reliability of "0.8," but since it has already been detected as a target organ, its reliability as a candidate related organ is "0." Therefore, the reliability of the candidate related organs, i.e., the "liver," "lungs," "lymph nodes," and "stomach," are "0," "0.6" (here, the maximum value between 0.5 and 0.6 is taken), "0.4," and "0.1," respectively (step S202 in FIG. 11).

[0110] Next, the identifying unit 1023 identifies the relevant organ as "lung" (step S203 in FIG. 11).

[0111] Next, the control unit 103 determines that there is a relevant organ (step S300 in FIG. 11: NO).

[0112] Next, the control unit 103 specifies the relevant organ "lungs" as a new target organ (step S300 in FIG. 11), and the processing of the analysis device 100C returns to step S101.

[0113] Up to this point, analysis device 100C has completed the circulation process indicated by "2" in FIG.

[0114] Next, analysis device 100C starts processing the circulation indicated by "3", as shown in Fig. 12. Analysis device 100C performs processing in the circulation indicated by "3" in the same manner as the circulation indicated by "2", identifies "lymph node" as a related organ, sets "lymph node" as a new target organ, and the processing of analysis device 100C returns to step S101.

[0115] Next, analysis device 100C starts the process of the loop indicated by "4" as shown in FIG. 12, and in the loop indicated by "4", the following process is performed.

[0116] First, the analysis device 100C determines a target organ image range of a "lymph node" that is a new target organ, based on a medical image of a subject, by the target organ image range determining unit 1011 (step S101 in FIG. 11).

[0117] Next, the target organ lesion detection unit 1012 detected three lesions in the target organ "lymph node" (the "lymph node" part is indicated by dots from the top to the center of Figure 12), and obtained the associated organ information of "Lesion 1 (lesion location: #####33#); associated organs: liver (0.8), lung (0.4)", "Lesion 2 (lesion location: %%%%%%%%); associated organs: lung (0.6), stomach (0.05)", and "Lesion 3 (lesion location: %%%%%%%%%%); associated organs: lung (0.5), lymph node (0.5), stomach (0.1)" (step S102 in Figure 11).

[0118] Next, the lesion analysis unit 1013 performs a lesion analysis on all the lesions in the "lymph node" and obtains lesion analysis information (step S103 in FIG. 11).

[0119] Next, the candidate associated organ identifying unit 1021 determines that all the candidate associated organs are the "liver", the "lung", the "lymph node", and the "stomach" (step S201 in FIG. 11).

[0120] Next, the reliability calculation unit 1022 calculates the reliability of each of the candidate related organs "liver", "lungs", "lymph nodes", and "stomach" for all the candidate related organs. Among the candidate related organs, the "liver", "lungs", and "lymph nodes" have already been detected as target organs, and therefore their reliability as candidate related organs is "0". Therefore, the reliability of the candidate related organs "liver", "lungs", "lymph nodes", and "stomach" is "0", "0", "0", and "0.1", respectively (step S202 in FIG. 11).

[0121] Next, the identification unit 1023 does not consider the "stomach" to be a related organ because the reliability of the "stomach" as a related organ is only "0.1", which is lower than the predetermined lower limit value "0.5" of the reliability of related organs that was set in advance, and identifies that there is no related organ (step S203 in Figure 11).

[0122] Next, the control unit 103 determines that no relevant organ has been identified (step S300 in FIG. 11: YES), and the processing of the analysis device 100C ends.

[0123] As described above, in the analysis device 100C according to the fourth embodiment, the user only needs to specify the "colon" as the target organ, and the analysis device 100C automatically performs metastatic cancer analysis on the "colon", the "liver" which is an organ related to the "colon", the "lung" which is an organ related to the "liver", and the "lymph node" which is an organ related to the "lung". As a result, the analysis device 100C according to the fourth embodiment can improve the detection accuracy of metastatic cancer. In addition, the analysis device 100C according to the fourth embodiment does not execute a corresponding fully automatic lesion detection and segmentation algorithm for all organs (including organs without lesions), but executes a fully automatic lesion detection and segmentation algorithm only for the related organs, thereby avoiding unnecessary and unnecessary calculations, avoiding the accumulation of errors, and reducing false positive results. In addition, the analysis device 100C according to the fourth embodiment can reduce the process of doctors manually reconfirming these false positives, thereby reducing the doctor's workload. Furthermore, in the analysis device 100C according to the fourth embodiment, unlike the conventional technology, the doctor does not need to manually check the relevant organs, and the burden on the doctor can be reduced. Therefore, the analysis device 100C according to the fourth embodiment can improve the efficiency of medical treatment.

[0124] (Modification) Although the embodiments have been described above, the present invention is not limited to the above-described embodiments and modifications.

[0125] (Variation 1) As described above, in the present embodiment, an example has been described in which lesion detection is performed on a target organ based on a trained deep learning model (first deep learning model) to obtain related organ information corresponding to the lesion, but the present invention is not limited thereto. For example, in the analysis device according to the first modification, related organ information including at least the name of a candidate related organ that is a candidate of a related organ related to the target organ and the reliability of the candidate related organ may be obtained based on a deep learning model without performing lesion detection on the target organ. This deep learning model is a model trained by training data so as to output at least the name of the candidate related organ and the reliability of the candidate related organ for the target organ. Here, the deep learning model in the first modification is an example of a "second deep learning model".

[0126] Furthermore, in the analysis device of variant example 1, the two types of deep learning models (first and second deep learning models) described above may be applied in a comprehensive manner to determine the names of candidate related organs and the reliability of the candidate related organs, thereby further improving the accuracy of metastatic cancer analysis.

[0127] (Variation 2) As described above, in the present embodiment, the analysis device determines the target organ image range based on the user's input, but the present invention is not limited to this.

[0128] For example, in the analysis device according to the second modification, the analysis device may identify the target organ in an automatic manner, without user input, by automatically selecting the range of the initial target organ based on the basic medical image information.

[0129] Here, the basic medical image information includes "Study Description", "Series Description", "Convolution Kernel", and the like.

[0130] For example, it is assumed that basic information of a medical image is as follows. ``Study Description:CT CHEST ABDOMEN PELVI Series Description:Lung+N14 A30 Convolution Kernel: Lungs Others"

[0131] In this case, the analysis device can select the lungs as the initial target organ based on the character string information "lungs" extracted in "Study Description" and "Convolution Kernel."

[0132] Also, it is assumed that basic information of a medical image is as follows. ``Study Description:Abdomen^2_1__AbdMultiPhaseCE (Adult) Series Description:Venous Phase 1.0 B30f Body Part Examined:ABDOMEN Others"

[0133] In this case, the analysis device cannot obtain the initial target organ from the string information alone, but can automatically select the liver as the initial target organ based on the extracted string “ABDOMEN” and the user configuration / software configuration.

[0134] In this way, the analysis device according to the second modification can automatically perform a rapid, accurate, and reliable metastatic cancer analysis on medical images of a subject without requiring any input from the user.

[0135] (Variation 3) As described above, in the fourth embodiment, an example has been described in which the analysis device automatically performs metastatic cancer analysis on a plurality of organs and their associated organs through multiple circulations, but the present invention is not limited to this.

[0136] For example, in the analysis device of variant example 3, rather than cyclically processing one identified related organ as a new target organ based on the names and respective reliability of multiple candidate related organs in the target organ, existing analysis modules corresponding to multiple candidate related organs may be directly applied to each of the multiple candidate related organs in the target organ for analysis.

[0137] In this case, the analysis device according to the third modification does not perform metastatic cancer analysis on all organs as in the conventional technology, but excludes unrelated organs based on the target organ, candidate related organs and their reliability, and performs analysis based on the existing analysis module on only the selected candidate related organs. This eliminates the need to execute a fully automatic lesion detection and segmentation algorithm for all organs (including organs without lesions), thereby reducing the accumulation of errors, reducing the number of false positives, and reducing the number of lesions that doctors need to manually recheck for these false positives, thereby reducing the workload of users such as doctors.

[0138] (Variation 4) As described above, an example has been described in which the analysis device trains a deep learning model using test images of a sample and corresponding diagnostic information from a doctor, and multiple follow-up test images of the subject and corresponding diagnostic information from a doctor as training data (e.g., Table 1), but this is not limited to this.

[0139] [Table 1]

[0140] For example, in the analysis device according to the fourth modification, a deep learning model may be trained using as training data (Table 2) information including a test image of a specimen and the corresponding diagnostic information of a doctor, multiple follow-up test images of the subject and the corresponding diagnostic information of a doctor, as well as biometric information and multimodal information. Specifically, the training data shown in Table 2 includes multiple follow-up test images of the subject and the corresponding diagnostic information of a doctor, as well as biometric information and multimodal information. The multimodal information is multiple types of information such as text, image, audio, and video, and is processed by multimodal AI, which is an AI (Artificial Intelligence) technology. In the example shown in Table 2, the multimodal information includes information such as an ultrasound image, which is a color image, and text such as "blood vessels inside," and the biometric information includes information that a biopsy test was performed and found to be malignant.

[0141] [Table 2]

[0142] As explained above, in the analysis device of variant example 4, by introducing additional information, it is possible to train a deep learning model that is closer to the actual situation, which is beneficial for improving the analysis accuracy rate of the analysis device.

[0143] As described above, the example of identifying related organs mainly considering lesions of organs has been described, but the present invention is not limited to this. For example, in this embodiment, related organs can be identified based on related information of anatomical structures (lymph nodes, etc.). This can contribute to improving the analysis accuracy of the analysis device.

[0144] In the present embodiment, the numerical ranges are merely examples, and can be appropriately set or changed according to actual needs.

[0145] In addition, although this embodiment describes in detail specific examples of an analysis device and an analysis method executed by it, this embodiment is not limited to this and can be realized in various forms including a medical image processing device, an integrated circuit, a program, and a medium on which a program is recorded, etc.

[0146] For example, the medical image processing apparatus according to this embodiment includes an image acquisition device that acquires a medical image including at least a target organ of a subject, and the above-mentioned analysis device.

[0147] Here, the image acquisition device may be a medical image collection device such as an ultrasound diagnostic device, an X-ray CT device, an MRI device, etc. Also, the image acquisition device may be a device that has a communication function and communicates with a medical image collection device such as an ultrasound diagnostic device, an X-ray CT device, an MRI device, etc., to acquire medical images. Furthermore, the image acquisition device may be a device that acquires medical images by reading a storage device that previously stores medical images of a target organ.

[0148] According to at least one of the embodiments described above, it is possible to improve the accuracy rate of detection of metastatic cancer and the efficiency of medical treatment.

[0149] Although some embodiments have been described, these embodiments are presented as examples 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 modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope of the invention and its equivalents as described in the claims, as well as in the scope and spirit of the invention. [Explanation of symbols]

[0150] 100 Analyzer 101 Target Organ Analysis Division 102 Related Organ Identification Department

Claims

1. a target organ analysis unit that performs an analysis based on a medical image of a target organ and acquires relevant organ information including at least a name of a candidate relevant organ that is a candidate of a relevant organ related to the target organ and a reliability of the candidate relevant organ; a related organ identification unit that identifies a related organ related to the target organ from the reliability of the candidate related organ based on the related organ information; An analysis device comprising:

2. a control unit that performs control so as to repeatedly perform a process of analyzing the relevant organ identified by the relevant organ identification unit as a new target organ in the target organ analysis unit, and then identifying a next new relevant organ by the relevant organ identification unit, until the relevant organ identification unit identifies that there is no relevant organ; The analysis device of claim 1 further comprising:

3. The target organ analysis unit includes: A target organ image range determination unit that automatically or manually determines a target organ image range for the medical image of the target organ; a target organ lesion detection unit that detects all lesions of the target organ within the target organ image range and acquires, for all the lesions, lesion position information for each lesion and the associated organ information corresponding to each lesion; The analysis device according to claim 1 or 2, comprising:

4. The target organ analysis unit includes: a lesion analysis unit that performs an analysis on all of the lesions based on the lesion position information and acquires lesion analysis information for each lesion; The analysis device according to claim 3 , further comprising:

5. The target organ lesion detection unit detects all lesions in the target organ based on a first deep learning model, and acquires lesion position information for each lesion and the associated organ information corresponding to each lesion. The analysis device according to claim 3.

6. The first deep learning model is a model trained with first training data so as to detect all lesions in the target organ for each organ, and output, for all the lesions, lesion position information for each lesion and the associated organ information corresponding to each lesion. The analysis device according to claim 5.

7. The first training data includes an examination image of a subject and corresponding diagnosis information of a doctor, and a plurality of follow-up examination images of the subject and corresponding diagnosis information of a doctor. The analysis device according to claim 6.

8. The first training data includes an examination image of a subject and corresponding diagnosis information of a physician, a plurality of follow-up examination images of the subject and corresponding diagnosis information of a physician, biometric information, and multimodal information. The analysis device according to claim 6.

9. The relevant organ identification unit includes: a candidate related organ identifying unit that identifies all of the candidate related organs based on the related organ information; a reliability calculation unit that calculates a reliability of each of the candidate related organs for all of the candidate related organs; an identification unit that identifies, as a related organ, the candidate related organ having the highest reliability or the candidate related organ having a reliability equal to or greater than a predetermined threshold; The analysis device according to claim 1 or 2, comprising:

10. The target organ analysis unit acquires the relevant organ information including at least a name of a candidate relevant organ that is a candidate of a relevant organ related to the target organ and a reliability of the candidate relevant organ based on a second deep learning model. The analysis device according to claim 1 or 2.

11. The second deep learning model is a model trained to output at least the name of the candidate associated organ and the confidence level of the candidate associated organ for the target organ. The analysis device according to claim 10.

12. a target organ analysis step of performing an analysis based on a medical image of a target organ to obtain relevant organ information including at least a name of a candidate relevant organ that is a candidate of a relevant organ related to the target organ and a reliability of the candidate relevant organ; a related organ identifying step of identifying a related organ related to the target organ from the reliability of the candidate related organ based on the related organ information; Analysis methods including:

13. a control step of controlling to repeatedly perform a process of analyzing the relevant organ identified in the relevant organ identification step as a new target organ in the target organ analysis step until the relevant organ is identified as no relevant organ in the relevant organ identification step, and then identifying a next new relevant organ in the relevant organ identification step; The method of claim 12 further comprising:

14. an image acquisition device for acquiring a medical image including at least a target organ of a subject; The analysis device according to claim 1 or 2, A medical image processing device comprising:

Images