Image processing device, image processing method, and image processing program

Abstract

To provide an image processing device, an image processing method, and an image processing program that enable appropriate presentation of a possibility that there is a lesion that is invisible in a medical image.SOLUTION: An image processing device derives a possibility that there is a lesion in a medical image and a degree of visibility indicating a degree of easiness to view of the lesion in the medical image. When the possibility indicates that there is a lesion in the medical image and the degree of visibility indicates that the lesion is invisible in the medical image, the image processing device issues a notification of information indicating that there is a lesion invisible in the medical image.SELECTED DRAWING: Figure 7

Description

Technical Field

[0001] The present disclosure relates to an image processing apparatus, an image processing method, and an image processing program.

Background Art

[0002] Patent Document 1 discloses a technique that determines a priority level for a lesion detection region in a medical image on which detection processing for a plurality of types of lesions has been performed, based on predetermined conditions, and changes the display mode of the detection region detected in the medical image according to the determined priority level.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In a medical image obtained by imaging a diagnosis target site of a patient having a lesion, the lesion may not be visualized in the medical image, or may be only faintly visualized. In this case, even if the reader of the medical image cannot visually recognize the lesion itself, the reader determines the presence or absence of the lesion based on changes in the shape and properties of surrounding tissues accompanying the occurrence of the lesion. Therefore, in this case, in order to effectively support the reader, it is preferable to appropriately present to the reader that a lesion that cannot be visually recognized may exist in the medical image.

[0005] The present disclosure has been made in view of the above circumstances, and an object thereof is to provide an image processing apparatus, an image processing method, and an image processing program capable of appropriately presenting that a lesion that cannot be visually recognized may exist in a medical image.

Means for Solving the Problems

[0006] The first embodiment of the image processing apparatus is an image processing apparatus comprising at least one processor, the processor derives the probability of a lesion being present in a medical image and the visibility score representing how easily the lesion is visible in the medical image, and when the probability indicates that a lesion is present in the medical image and the visibility score indicates that the lesion is not visible in the medical image, the processor provides information indicating that there is a lesion that is not visible in the medical image.

[0007] The second embodiment of the image processing apparatus further controls the processor to display different aspects depending on the case in which the probability indicates that a lesion is present in the medical image and visibility indicates that the lesion is not visible in the medical image; the probability indicates that a lesion is present in the medical image and visibility indicates that the lesion is visible in the medical image; and the probability indicates that a lesion is not present in the medical image.

[0008] The third embodiment of the image processing apparatus is an image processing apparatus of the first or second embodiment in which the processor takes a medical image as input and outputs a probability, and derives a probability by inputting a medical image to be diagnosed to a first pre-trained model which has been trained using multiple sets of medical images and probabilities as training data.

[0009] The fourth embodiment of the image processing apparatus is an image processing apparatus according to any one of the first to third embodiments, wherein the processor takes a medical image as input and outputs visibility, and derives visibility by inputting a medical image to be diagnosed to a second pre-trained model which has been trained using multiple sets of medical images and visibility as training data.

[0010] The fifth embodiment of the image processing apparatus is an image processing apparatus of the fourth embodiment in which, if the probability indicates that a lesion is present in the medical image, the processor derives visibility by inputting the region portion of the medical image in which a lesion is determined to be present into a second trained model.

[0011] The sixth embodiment of the image processing apparatus is an image processing apparatus of the first or second embodiment in which the processor derives a probability based on one or more indirect findings in the medical image.

[0012] The seventh embodiment of the image processing apparatus is an image processing apparatus of the sixth embodiment in which the processor derives a first probability based on one or more indirect findings, derives a second probability by performing lesion detection processing on the medical image, derives a third probability based on the first and second probabilities, and when the third probability indicates that a lesion is present in the medical image and the visibility indicates that the lesion is not visible in the medical image, the processor reports information indicating that there is a lesion that is not visible in the medical image.

[0013] The eighth aspect of the image processing method involves a processor in an image processing device that derives the probability of a lesion being present in a medical image and a visibility score representing how easily the lesion is visible in the medical image, and when the probability indicates that a lesion is present in the medical image and the visibility score indicates that the lesion is not visible in the medical image, the processor performs a process to notify the system of information indicating that there is a lesion that is not visible in the medical image.

[0014] The ninth embodiment of the image processing program derives the probability of a lesion being present in a medical image and the visibility score, which represents how easily the lesion is visible in the medical image. If the probability indicates that a lesion is present in the medical image and the visibility score indicates that the lesion is not visible in the medical image, the program causes the processor of the image processing device to perform a process that notifies information indicating that there is a lesion that is not visible in the medical image. [Effects of the Invention]

[0015] This disclosure allows for the appropriate presentation of the possibility that lesions not visible in medical images may exist. [Brief explanation of the drawing]

[0016] [Figure 1] This is a block diagram illustrating the schematic configuration of a medical information system. [Figure 2] This is a block diagram showing an example of the hardware configuration of an image processing apparatus. [Figure 3] This is a block diagram showing an example of the functional configuration of the image processing apparatus. [Figure 4] This is a diagram for explaining the first trained model. [Figure 5] This is a diagram showing an example of a region determined to have a lesion present. [Figure 6] This is a diagram for explaining the second trained model. [Figure 7] This is a diagram showing an example of a display screen. [Figure 8] This is a flowchart showing an example of diagnosis support processing. [Figure 9] This is a diagram for explaining processing of deriving the probability that a lesion is present from indirect findings according to a modification. [Figure 10] This is a diagram for explaining processing of deriving the probability that a lesion is present based on indirect findings and lesion detection processing according to a modification. [Figure 11] This is a diagram for explaining the second trained model according to a modification. [Description of Embodiments]

[0017] Hereinafter, exemplary embodiments for implementing the technology of the present disclosure will be described in detail with reference to the drawings.

[0018] First, the configuration of a medical information system 1 according to the present embodiment will be described with reference to FIG. 1. As shown in FIG. 1, the medical information system 1 includes an image processing apparatus 10, an imaging apparatus 12, and an image storage server 14. The image processing apparatus 10, the imaging apparatus 12, and the image storage server 14 are communicatively connected to each other via a wired or wireless network 18. The image processing apparatus 10 is, for example, a computer such as a personal computer or a server computer.

[0019] The imaging device 12 is a device that generates a medical image representing a part of a subject to be diagnosed by imaging that part of the subject. Examples of imaging devices 12 include plain X-ray imaging devices, endoscopes, CT (Computed Tomography) devices, MRI (Magnetic Resonance Imaging) devices, and PET (Positron Emission Tomography) devices. In this embodiment, an example is described in which the imaging device 12 is a CT device and the part to be diagnosed is the abdomen. That is, the imaging device 12 according to this embodiment generates a CT image of the subject's abdomen as a three-dimensional medical image consisting of multiple tomographic images. The medical image generated by the imaging device 12 is transmitted to the image storage server 14 via the network 18 and stored by the image storage server 14.

[0020] The image storage server 14 is a computer that stores and manages various types of data, and is equipped with a large-capacity external storage device and database management software. The image storage server 14 receives medical images generated by the imaging device 12 via the network 18, and stores and manages the received medical images. The storage format of image data by the image storage server 14 and communication with other devices via the network 18 are based on protocols such as DICOM (Digital Imaging and Communication in Medicine).

[0021] Next, the hardware configuration of the image processing device 10 according to this embodiment will be described with reference to Figure 2. As shown in Figure 2, the image processing device 10 includes a CPU (Central Processing Unit) 20, a memory 21 as a temporary storage area, and a non-volatile storage unit 22. The image processing device 10 also includes a display 23 such as a liquid crystal display, input devices 24 such as a keyboard and mouse, and a network interface 25 connected to a network 18. The CPU 20, memory 21, storage unit 22, display 23, input devices 24, and network interface 25 are connected to a bus 27. The CPU 20 is an example of a processor according to the disclosed technology.

[0022] The storage unit 22 is implemented by an HDD (Hard Disk Drive), SSD (Solid State Drive), or flash memory, etc. The image processing program 30 is stored in the storage unit 22 as a storage medium. The CPU 20 reads the image processing program 30 from the storage unit 22, expands it into memory 21, and executes the expanded image processing program 30.

[0023] Furthermore, the memory unit 22 stores the trained model 32 and the trained model 34. Details of the trained model 32 and the trained model 34 will be described later.

[0024] Incidentally, in medical images obtained by imaging the diagnostic area of ​​a patient with a lesion, the lesion may not be visible in the medical image, or it may be only slightly visible. In this case, even if the lesion itself cannot be visually observed, the radiologist or other image interpreter will determine the presence or absence of a lesion based on indirect findings associated with the occurrence of the lesion. In this embodiment, a lesion refers to an area of ​​abnormality that is directly treated, such as cancer, cysts, and inflammation. In this embodiment, indirect findings refer to findings that represent at least one of the characteristics of the shape and properties of the surrounding tissue of the lesion. For example, in the pancreatic region of an abdominal CT image, even if pancreatic cancer cannot be visually observed, the radiologist may suspect the occurrence of pancreatic cancer based on an example of indirect findings, such as dilation of the pancreatic duct.

[0025] The image processing device 10 according to this embodiment has a function to inform the radiologist that there may be lesions in the medical image that are not visible to the viewer, in order to effectively support the radiologist.

[0026] Next, with reference to Figure 3, the functional configuration of the image processing apparatus 10 according to this embodiment will be described. As shown in Figure 3, the image processing apparatus 10 includes an acquisition unit 40, a first output unit 42, a second output unit 44, and a display control unit 46. The CPU 20 executes the image processing program 30, thereby enabling the acquisition unit 40, the first output unit 42, the second output unit 44, and the display control unit 46 to function.

[0027] The acquisition unit 40 acquires the medical image to be diagnosed (hereinafter referred to as "diagnostic image") from the image storage server 14 via the network interface 25.

[0028] The first derivation unit 42 uses the trained model 32 to derive the probability that a lesion exists in the diagnostic image acquired by the acquisition unit 40 (hereinafter referred to as "probability of lesion presence"). The trained model 32 is a trained model that takes a medical image as input and outputs the probability of lesion presence, and is a trained model that has been trained using multiple sets of medical images and the probability of lesion presence in those medical images as training data (sometimes referred to as training data). The trained model 32 is, for example, composed of a CNN (Convolutional Neural Network). In this embodiment, the trained model 32 outputs either "lesion present" indicating that a lesion is present or "no lesion" indicating that a lesion is not present as the probability of lesion presence. The trained model 32 is an example of the first trained model relating to the disclosed technology.

[0029] As shown in Figure 4, the first derivation unit 42 receives the diagnostic target image acquired by the acquisition unit 40 as input to the trained model 32. The trained model 32 outputs the probability of lesion presence corresponding to the input diagnostic target image. Based on this, the first derivation unit 42 derives the probability of lesion presence.

[0030] The trained model 32 may output a numerical value between 0 and 1. In this case, the first derivation unit 42 may determine that there is a lesion if the output of the trained model 32 is equal to or greater than a threshold (for example, 0.5), and determine that there is no lesion if it is less than the threshold.

[0031] Furthermore, as shown in Figure 5, the first derivation unit 42 identifies the region in the image to be diagnosed where a lesion is determined to exist, if the probability of lesion presence output from the trained model 32 is "lesion present". Specifically, the first derivation unit 42 identifies regions where the contribution rate when the trained model 32 outputs "lesion present" is above a certain value as the region in the image to be diagnosed where a lesion is determined to exist. In this case, the contribution rate can be calculated using known techniques such as Grad-CAM (Gradient-weighted Class Activation Mapping). In the example in Figure 5, the region inside the dashed line, that is, the head of the pancreas and the surrounding region, is shown as an example where a lesion was determined to exist.

[0032] The second derivation unit 44 uses the trained model 34 to derive a visibility score (hereinafter referred to as "lesion visibility score") that represents how easily a lesion can be seen in the diagnostic image acquired by the acquisition unit 40. The trained model 34 is a trained model that takes a medical image as input and outputs a lesion visibility score, and is a trained model that has been trained using multiple sets of medical images and the lesion visibility scores in those medical images as training data. The trained model 34 is composed of, for example, a CNN. The trained model 34 according to this embodiment outputs either "visible," which indicates that the lesion can be seen, or "invisible," which indicates that the lesion cannot be seen, as the lesion visibility score. The trained model 34 is an example of a second trained model relating to the disclosed technology.

[0033] As shown in Figure 6, the second derivation unit 44 inputs the region in the diagnostic target image identified by the first derivation unit 42 in which a lesion is determined to exist to the trained model 34. The trained model 34 outputs the lesion visibility score corresponding to the input region. In this way, the second derivation unit 44 derives the lesion visibility score.

[0034] The second derivation unit 44 may output a numerical value between 0 and 1. In this case, the second derivation unit 44 may determine that the output of the trained model 34 is "visible" if it is equal to or greater than a threshold (for example, 0.5), and determine that it is "not visible" if it is less than the threshold.

[0035] The display control unit 46 notifies information indicating that an invisible lesion exists in the image to be diagnosed when the probability of lesion presence derived by the first derivation unit 42 is "lesion present" and the visibility of the lesion derived by the second derivation unit 44 is "not visible". Specifically, in this case, as shown in Figure 7 as an example, the display control unit 46 controls the display 23 to display the image to be diagnosed, information representing the area in the image to be diagnosed in which a lesion is determined to exist as identified by the first derivation unit 42, and a message indicating that there may be an invisible lesion in the image to be diagnosed. In the example in Figure 7, the outline of the area in the image to be diagnosed in which a lesion is determined to exist as identified by the first derivation unit 42 is shown with a dashed line.

[0036] The display control unit 46 may perform control to display in a different manner in each of the following cases (1), (2), and (3). (1) When the probability of lesion presence is "lesion present" and the degree of lesion visibility is "not visible" (2) When the probability of lesion presence is "lesion present" and the degree of lesion visibility is "visible" (3) When the probability of lesion presence is "no lesion"

[0037] In this case, the display control unit 46 controls the following on the display 23: (1) as described above, the image to be diagnosed, information representing the region in the image to be diagnosed identified by the first derivation unit 42 in which a lesion is determined to exist, and a message indicating that there may be a lesion in the image to be diagnosed that is not visible. In this case, (2) the display control unit 46 controls the display 23 to display the image to be diagnosed and information representing the region in the image to be diagnosed identified by the first derivation unit 42 in which a lesion is determined to exist. In this case, (3) the display control unit 46 controls the display 23 to display the image to be diagnosed and a message indicating that there is a high probability that no lesion exists. In this way, by controlling the display control unit 46 to display each of the cases (1) to (3) in a distinguishable manner, the radiologist can easily grasp the presence or absence of a lesion and whether or not the lesion is visible.

[0038] Furthermore, in this embodiment, the display control unit 46 may distinguish lesion visibility by changing the type, thickness, or color of the outline lines of the area in the diagnostic image where a lesion is determined to be present, according to the lesion visibility level. Alternatively, the display control unit 46 may distinguish lesion visibility levels by controlling the display of a mark pre-assigned to each lesion visibility level near the area where a lesion is determined to be present. In addition, if the lesion visibility level is a numerical value, the display control unit 46 may control the display to show that numerical value on the display 23.

[0039] Furthermore, the display control unit 46 may fill in the area in the diagnostic target image identified by the first derivation unit 42 where a lesion is determined to exist with a pre-set color. Also, the display control unit 46 does not need to display information representing the area where the lesion visibility is "visible".

[0040] Next, the operation of the image processing apparatus 10 according to this embodiment will be explained with reference to Figure 8. The diagnostic support process shown in Figure 8 is executed when the CPU 20 executes the image processing program 30. The diagnostic support process shown in Figure 8 is executed, for example, when a user inputs an instruction to start execution.

[0041] In step S10 of Figure 8, the acquisition unit 40 acquires the image to be diagnosed from the image storage server 14 via the network interface 25. In step S12, the first derivation unit 42, as described above, uses the trained model 32 to derive the probability of lesion presence in the image to be diagnosed acquired in step S10. In step S14, the first derivation unit 42 determines whether the probability of lesion presence derived in step S12 is "lesion present". If this determination is positive, the process proceeds to step S16.

[0042] In step S16, the first derivation unit 42 identifies the region in the diagnostic image acquired in step S10 in which a lesion is determined to be present, as described above. In step S18, the second derivation unit 44 uses the trained model 34 to derive the degree of lesion visibility in the diagnostic image acquired in step S10, as described above.

[0043] In step S20, the display control unit 46 determines whether the lesion visibility derived in step S18 is "not visible". If this determination is negative, the process moves to step S24; if it is positive, the process moves to step S22. In step S22, the display control unit 46 controls the display 23 to display the image to be diagnosed, information representing the region identified in step S16, and a message indicating that there may be a lesion that is not visible in the image to be diagnosed. When the process in step S22 is completed, the diagnostic support process ends. In step S24, the display control unit 46 controls the display 23 to display the image to be diagnosed and information representing the region identified in step S16. When the process in step S24 is completed, the diagnostic support process ends.

[0044] If the determination in step S14 is negative, the process proceeds to step S26. In step S26, the display control unit 46 controls the display 23 to display a message indicating that there is a high probability that the image to be diagnosed and the lesion are not present. When the process in step S26 is completed, the diagnostic support process ends.

[0045] As described above, this embodiment makes it possible to appropriately indicate that there may be lesions that are not visible in medical images.

[0046] In the above embodiment, the first derivation unit 42 has been described as deriving the probability of lesion presence using a trained model 32, but it is not limited to this. As an example, as shown in Figure 9, the first derivation unit 42 may derive the probability of lesion presence based on one or more indirect findings in the diagnostic image. In the example in Figure 9, the first derivation unit 42 derives "lesion present" as the probability of lesion presence based on two indirect findings: enlargement in the head of the pancreas and atrophy in the tail of the pancreas. In this case, cancer is suspected to be present in the pancreas. In this case, the first derivation unit 42 may derive the probability of lesion presence using a lookup table in which one or more indirect findings are associated with lesion names.

[0047] In this case, the first derivation unit 42 may derive indirect findings using a trained model that takes medical images as input and outputs information representing indirect findings, the trained model being trained using multiple sets of medical images and information representing indirect findings present in those medical images as training data.

[0048] Furthermore, as shown in Figure 10 as an example, the first derivation unit 42 may derive a first probability of lesion presence based on one or more indirect findings in the image to be diagnosed, and derive a second probability of lesion presence by performing a lesion detection process on the image to be diagnosed. In this case, the lesion detection process may be a process using the trained model 32, as in the above embodiment, or a process using a known lesion detection algorithm. In this case, the first derivation unit 42 may also derive a third probability of lesion presence based on the first and second probabilities of lesion presence. For example, if at least one of the first and second probabilities of lesion presence is "lesion present", the first derivation unit 42 derives "lesion present" as the third probability of lesion presence. Also, if both the first and second probabilities of lesion presence are "no lesion", the first derivation unit 42 derives "no lesion" as the third probability of lesion presence. In this case, the display control unit 46 may notify information indicating that an invisible lesion exists in the diagnostic image when the third probability of lesion presence is "lesion present" and the lesion visibility is "not visible". In the example in Figure 10, the area inside the dashed line represents the area in which a lesion detected by the lesion detection process is determined to exist.

[0049] In this case, the display control unit 46 may also perform control to distinguish between the portion of the image to be diagnosed that is determined to have a lesion by indirect findings and the portion that is determined to have a lesion by lesion detection processing.

[0050] Furthermore, in the above embodiment, the second derivation unit 44 was described as deriving the lesion visibility by inputting the region portion in the diagnostic target image identified by the first derivation unit 42 into the trained model 34, in which it is determined that a lesion exists. However, the invention is not limited to this. As an example, as shown in Figure 11, the second derivation unit 44 may also derive the lesion visibility by inputting the same image (i.e., the diagnostic target image) into the trained model 34 as the image input to the trained model 32.

[0051] Furthermore, in the above embodiment, the second derivation unit 44 was described as deriving the lesion visibility in the diagnostic image using a trained model 34, but it is not limited to this. The second derivation unit 44 may also derive the lesion visibility based on the variance of the CT values ​​and the difference in the mean values ​​of the CT values ​​with respect to the surrounding area.

[0052] Furthermore, although the above embodiment describes a case where the trained model 32 and trained model 34 are composed of CNNs, the embodiment is not limited to this. The trained model 32 and trained model 34 may be composed of machine learning methods other than CNNs.

[0053] Furthermore, although the above embodiment described the case in which CT images were used as the diagnostic target image, the invention is not limited to this. Other medical images besides CT images, such as radiographic images taken with a simple X-ray imaging device and MRI images taken with an MRI device, may also be used as the diagnostic target image.

[0054] Furthermore, the processes in steps S10 to S12 of the diagnostic support process according to the above embodiment may be executed before the user inputs an instruction to start execution. In this case, when the user inputs an instruction to start execution, steps S14 onwards are executed and the screen display is performed.

[0055] Furthermore, in the above embodiment, the hardware structure of the processing unit that executes various processes, such as the acquisition unit 40, the first derivation unit 42, the second derivation unit 44, and the display control unit 46, can be the various processors shown below. As mentioned above, these various processors include a CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, as well as programmable logic devices (PLDs), such as FPGAs (Field Programmable Gate Arrays), which are processors whose circuit configuration can be changed after manufacturing, and dedicated electrical circuits, such as ASICs (Application Specific Integrated Circuits), which are processors with circuit configurations specifically designed to execute specific processes.

[0056] A single processing unit may consist of one of these various processors, or it may consist of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, multiple processing units may be composed of a single processor.

[0057] Examples of configuring multiple processing units with a single processor include, firstly, a configuration where one or more CPUs and software combine to form a single processor, which then functions as multiple processing units, as exemplified by client and server computers. Secondly, a configuration using a processor that realizes the functions of the entire system, including multiple processing units, on a single IC (Integrated Circuit) chip, as exemplified by System on Chip (SoC). Thus, various processing units are configured, in terms of hardware structure, using one or more of the above-mentioned processors.

[0058] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits, which are combinations of circuit elements such as semiconductor devices.

[0059] Furthermore, although the above embodiment describes an embodiment in which the image processing program 30 is pre-stored (installed) in the storage unit 22, the invention is not limited to this. The image processing program 30 may be provided in the form of a recording medium such as a CD-ROM (Compact Disc Read Only Memory), DVD-ROM (Digital Versatile Disc Read Only Memory), or USB (Universal Serial Bus) memory. Alternatively, the image processing program 30 may be provided in the form of a download from an external device via a network. [Explanation of symbols]

[0060] 1. Medical Information System 10 Image Processing Device 12. Imaging device 14 Image storage server 18 Network 20 CPU 21 memory 22 Memory section 23 displays 24 Input devices 25 Network Interface 27 Bus 30 Image Processing Programs 32, 34 Pre-trained models 40 Acquisition Department 42 First derivation part 44 Second derivation part 46 Display Control Unit

Claims

1. An image processing apparatus comprising at least one processor, The aforementioned processor, Information indicating whether or not a lesion is present in a medical image, and visual confirmation / denial information indicating whether or not the radiologist can visually confirm the lesion in the medical image are derived. If the lesion presence information indicates that the lesion is present in the medical image, and the visual confirmation / non-confirmation information indicates that the radiologist cannot visually confirm the lesion in the medical image, then information indicating that there may be a lesion in the medical image that the radiologist cannot visually confirm is provided. Image processing device.

2. The aforementioned processor further, The lesion presence information indicates that the lesion is present in the medical image, and the visual confirmation / non-confirmation information indicates that the radiologist cannot visually confirm the lesion in the medical image. The lesion presence information indicates that the lesion is present in the medical image, and the visual confirmation / denial information indicates that the radiologist can visually confirm the lesion in the medical image; The information regarding the presence of the lesion indicates that the lesion is not present in the medical image, In each of these cases, control is performed to display the content in a different manner. The image processing apparatus according to claim 1.

3. The aforementioned processor, A trained model that takes the aforementioned medical images as input and outputs the lesion presence information, wherein the lesion presence information is derived by inputting the medical image of the target for diagnosis to a first trained model that has been trained using multiple sets of the aforementioned medical images and the lesion presence information as training data. The image processing apparatus according to claim 1 or claim 2.

4. The aforementioned processor, A trained model that takes the aforementioned medical image as input and outputs the aforementioned approval / disapproval information, wherein the approval / disapproval information is derived by inputting the medical image to be diagnosed to a second trained model that has been trained using multiple sets of the aforementioned medical images and the aforementioned approval / disapproval information as training data. The image processing apparatus according to claim 1 or claim 2.

5. The aforementioned processor, If the lesion presence information indicates that the lesion is present in the medical image, the visual approval / disapproval information is derived by inputting the region in the medical image in which the lesion is determined to be present into the second trained model. The image processing apparatus according to claim 4.

6. The aforementioned processor, Based on one or more indirect findings in the aforementioned medical image, the information regarding the presence of the lesion is derived. The image processing apparatus according to claim 1 or claim 2.

7. The aforementioned processor, Based on the one or more indirect findings, the first information regarding the presence of the lesion is derived. By performing a lesion detection process on the aforementioned medical image, a second lesion presence information is derived. Based on the first lesion presence information and the second lesion presence information, the third lesion presence information is derived. If the third lesion presence information indicates that the lesion is present in the medical image, and the visual confirmation / non-confirmation information indicates that the radiologist cannot visually confirm the lesion in the medical image, then information indicating that there is a possibility of a lesion present in the medical image that cannot be visually confirmed by the radiologist is provided. The image processing apparatus according to claim 6.

8. Information indicating whether or not a lesion is present in a medical image, and visual confirmation / denial information indicating whether or not the radiologist can visually confirm the lesion in the medical image are derived. If the lesion presence information indicates that the lesion is present in the medical image, and the visual confirmation / non-confirmation information indicates that the radiologist cannot visually confirm the lesion in the medical image, then information indicating that there may be a lesion in the medical image that the radiologist cannot visually confirm is provided. An image processing method in which the processing is performed by a processor in an image processing device.

9. Information indicating whether or not a lesion is present in a medical image, and visual confirmation / denial information indicating whether or not the radiologist can visually confirm the lesion in the medical image are derived. If the lesion presence information indicates that the lesion is present in the medical image, and the visual confirmation / non-confirmation information indicates that the radiologist cannot visually confirm the lesion in the medical image, then information indicating that there may be a lesion in the medical image that the radiologist cannot visually confirm is provided. An image processing program that causes the processor in an image processing device to perform the processing.

Images