Medical image processing system, method of operating a medical image processing system

Abstract

A medical image processing system and a working method of the medical image processing system that can specify an image most suitable for diagnosis while preventing overlooking of an image that should be observed. An endoscope system (10) sequentially acquires a plurality of medical images generated by continuously capturing an observation target, detects a region of interest by performing recognition processing on each of the plurality of medical images, and sequentially outputs and displays the plurality of medical images on a monitor (18). In a case where a designation operation is performed on a medical image being displayed, from among the medical image being displayed and a medical image captured on at least one of a side before the medical image being displayed and a side after the medical image being displayed, a medical image in which the region of interest is most visible is selected as a medical image designated by the designation operation.

Description

Technical Field

[0001] This invention relates to a medical image processing system and a method for operating the medical image processing system. Background Technology

[0002] In the medical field, medical images such as endoscopic images, ultrasound images, X-ray images, CT (Computed Tomography) images, and MR (Magnetic Resonance) images are used for patient diagnosis and observation. Based on these image diagnoses, doctors and others make decisions on treatment plans.

[0003] For example, Patent Document 1 describes a structure in which multiple medical images (endoscopic images) generated by continuously photographing the object being observed are sequentially output and displayed on a monitor, and the user (doctor) obtains the endoscopic image specified by operating the freeze switch while observing the monitor as a still image.

[0004] Furthermore, Patent Document 2 describes a structure in which an abnormal area (area of ​​interest) that should be observed is detected from each endoscope image by a learner, including lesions, and the endoscope image with the highest visibility of the area of ​​interest is set as the representative image from a series of endoscope images including common lesions.

[0005] Patent Document 1: Japanese Patent No. 4694255

[0006] Patent Document 2: Japanese Patent No. 6351323

[0007] However, in the aforementioned Patent Document 1, there is a problem that the image specified by the doctor may not be the most suitable image for diagnosis. That is, since the monitor screen is updated (images are captured) in units of tens of frames per second, sometimes the endoscopic image several frames before (or after) the doctor operates the freeze switch is a more visible and suitable image for diagnosis.

[0008] Furthermore, in the aforementioned Patent Document 2, since the representative image is mechanically set regardless of the doctor's intention, the doctor may miss some images if there are images to be observed before or after the representative image. Summary of the Invention

[0009] The present invention was made in view of the above background, and its object is to provide a medical image processing system that can specify the most suitable image for diagnosis while preventing the omission of images that should be observed, and a method for operating the medical image processing system.

[0010] To achieve the above objectives, the medical image processing system of the present invention includes a memory for storing program commands and a processor for executing the program commands. The processor performs the following processing: sequentially acquiring multiple medical images generated by continuously capturing images of an observed object; detecting a region of interest by performing recognition processing on each of the multiple medical images, and sequentially outputting the multiple medical images to a monitor for display; and when a specified operation for a specified medical image is performed on a medical image displayed on the monitor, selecting the medical image with the highest visibility of the region of interest from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor as the medical image specified by the specified operation.

[0011] The medical image with the highest contrast can be determined as having the highest visibility.

[0012] The region of interest can be identified and processed by processing, including the lesion site.

[0013] The identification process can include the identification of regions of interest.

[0014] You can select a specific medical image from the medical images captured within a specified period of time, which can be displayed on the monitor.

[0015] The specified operation can be a freeze operation that stops the monitor from updating the screen and continues to display a medical image.

[0016] The specified operation can be a capture operation that acquires a medical image displayed on a monitor as a recording image.

[0017] Convolutional Neural Networks can be used in recognition processing.

[0018] Medical images can be images obtained from an endoscope.

[0019] Furthermore, in order to achieve the above objectives, the medical image processing system of the present invention provides a method for operating the system, wherein the medical image processing system includes a memory for storing program commands and a processor for executing program commands, and the processor performs the following processing: sequentially acquiring multiple medical images generated by continuously capturing images of an observed object; detecting a region of interest by performing recognition processing on each of the multiple medical images, and sequentially outputting the multiple medical images to a monitor for display; and when a specified operation for a specified medical image is performed on the medical image displayed on the monitor, selecting the medical image with the highest visibility of the region of interest from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor as the medical image specified by the specified operation.

[0020] Invention Effects

[0021] According to the present invention, a medical image processing system and a method for operating the medical image processing system are provided that can specify the most suitable image for diagnosis while preventing the omission of images that should be observed. Attached Figure Description

[0022] Figure 1 This is an external view of the endoscope system.

[0023] Figure 2 It is a block diagram representing the function of an endoscope system.

[0024] Figure 3 It is a flowchart representing the action flow of the specified image selection section.

[0025] Figure 4 It is a flowchart representing the process of selecting and processing a specified image. Detailed Implementation

[0026] exist Figure 1 In this system, the endoscope system 10 (medical image processing system) includes an endoscope 12, a light source device 14, a processor device 16, a monitor 18, and a UI (User Interface) 19. The endoscope 12 is optically connected to the light source device 14 and electrically connected to the processor device 16.

[0027] The endoscope 12 has an insertion portion 12a that is inserted into the body of the object being observed, an operating portion 12b provided at the base of the insertion portion 12a, a bending portion 12c provided at the front end of the insertion portion 12a, and a front end portion 12d. The bending portion 12c is bent by operating the operating portion 12b. The front end portion 12d is oriented in a desired direction by the bending action of the bending portion 12c.

[0028] Furthermore, the operation unit 12b is equipped with: a designation switch 12g for designating a medical image (an endoscopic image in this embodiment); a mode switching switch 12h for switching observation modes; and a zoom operation unit 12i for changing the zoom magnification. The designated operations include a freeze operation to stop the monitor 18's screen update so that a single (frame) medical image continues to be displayed, and an acquisition (in storage device 58a (reference)) operation. Figure 2 The medical image displayed on the monitor 18 is saved in the record and captured as a still image. For example, for the medical image displayed on the monitor 18, a freeze operation is performed by long-pressing the designated switch 12g, and a capture operation is performed by short-pressing the designated switch 12g once.

[0029] Furthermore, in this embodiment, the freeze and capture operations are performed by a single operating component (designated switch 12g), but the freeze and capture operating components may also be provided separately. In this embodiment, the freeze and capture operating components are located in the operating unit 12b, but they may also be located in UI19. Moreover, in this embodiment, an example of an endoscope system capable of performing both freeze and capture operations has been described; however, the invention may also be applied to endoscope systems capable of performing only one of these operations.

[0030] The light source device 14 includes a light source unit 20 that emits illumination light for illuminating the object being observed (see reference). Figure 2 The illumination light from the light source unit 20 is guided by the light guide 25 (reference). Figure 2 The light is guided to illuminate the object being observed from the front end 12d. The object being observed, illuminated by the illumination light from the light source unit 20, is monitored by the camera sensor 44 (reference) built into the front end 12d. Figure 2 (Photographed)

[0031] The processor device 16 is electrically connected to the monitor 18 and the UI 19. The monitor 18 outputs and displays an image of the observed object, information accompanying the image, etc. The UI 19 has a keyboard, mouse, touchpad, microphone, etc., and has the function of accepting input operations such as function settings. In addition, an external memory (not shown) can be connected to the processor device 16.

[0032] exist Figure 2 In this device, the light source device 14 includes the aforementioned light source section 20. The light source section 20 is connected to the light source control section 21 of the processor device 16, and the emission spectrum and emission time of the illumination light emitted by the light source section 20 are controlled by the light source control section 21.

[0033] In this embodiment, the light source unit 20 emits ordinary light and special light with different emission spectra. Ordinary light is, for example, white light. White light includes, for example, violet light with a wavelength band of 380-420 nm, blue light with a wavelength band of 420-500 nm, green light with a wavelength band of 480-600 nm, and red light with a wavelength band of 600-650 nm. The endoscopic image (ordinary image) captured by illuminating the object under ordinary light is displayed on the monitor 18.

[0034] Specialized light can be of one type or multiple types. For example, violet light (peak wavelength 400nm–420nm), which has a high absorption coefficient due to hemoglobin in blood, emits more light than ordinary light. Endoscopic images (specialized images) captured by illuminating the subject with this specialized light are used to generate vascular images (live information images) that display the superficial vascular or glandular structures.

[0035] Furthermore, as a special light, for example, only the aforementioned purple light is emitted. Compared to the aforementioned case (where the amount of purple light emitted is greater than that of ordinary light), the endoscopic image (special image) captured by illuminating the observed object with this special light is used to generate a vascular image (live information image) that shows more superficial vascular or glandular structures.

[0036] In addition, a special light source is used, for example, blue-violet light (peak wavelength 470 nm to 480 nm) that emits light with a different absorption coefficient between oxidized and deoxidized hemoglobin. The endoscopic image (special image) taken by illuminating the subject with this special light is used to generate an oxygen saturation image (live information image) that shows the oxygen saturation of hemoglobin in the blood.

[0037] Furthermore, special light, such as the aforementioned violet light, blue-violet light, and red light (peak wavelength of 620nm to 630nm), emits a greater amount of light than ordinary light. Endoscopic images (special images) captured by illuminating the object with this special light are used to generate chromatic aberration images (live information images) that amplify the chromatic aberration between the lesion and areas outside the lesion. Moreover, the type of special light and the type of live information image generated from endoscopic images (special images) captured using various special lights are not limited to the types described above and can be appropriately varied.

[0038] Illumination light from the light source 20 is incident on the aforementioned light guide 25 via the light path coupling section 23, which is composed of a reflector, a lens, etc. The light guide 25 is built into the endoscope 12 and the universal plug (the plug connecting the endoscope 12 to the light source device 14 and the processor device 16). The light guide 25 propagates the light from the light path coupling section 23 to the front end portion 12d of the endoscope 12.

[0039] An illumination optical system 30a and an imaging optical system 30b are provided at the front end 12d of the endoscope 12. The illumination optical system 30a has an illumination lens 32, through which illumination light propagating via the light guide 25 illuminates the object being observed. The imaging optical system 30b has an objective lens 42, a zoom lens 43, and an image sensor 44. By illuminating the object with illumination light, light from the object being observed passes through the objective lens 42 and the zoom lens 43 and enters the image sensor 44. Thus, an image of the object being observed is formed on the image sensor 44. The zoom lens 43 is a lens used to magnify the object being observed, and it can be moved between a telephoto end and a wide-angle end by operating the zoom operation unit 12i.

[0040] The camera sensor 44 is a color sensor. In this embodiment, a primary color sensor is used, which has three types of pixels: B pixels with a B (blue) filter, G pixels with a G (green) filter, and R pixels with an R (red) filter. Such a camera sensor 44 can be a CCD (Charge-Coupled Device) or CMOS (Complementary Metal Oxide Semiconductor) camera sensor, etc.

[0041] The camera sensor 44 is controlled by the camera sensor control unit 45 of the processor device 16. The camera sensor control unit 45 performs image capture (signal readout of the camera sensor 44) at a predetermined period (e.g., 60 times per second). Then, along with the image capture, an image signal is output from the camera sensor 44. For example, in the case of image capture at a period of 60 times per second, an image signal of 60 frames per second is output.

[0042] Alternatively, instead of the primary color camera sensor 44 equipped with RGB primary color filters, a complementary color camera sensor with complementary color filters for C (cyan), M (magenta), Y (yellow), and G (green) can be used. When using a complementary color camera sensor, a CMYG four-color image signal is output. Therefore, by converting the complementary color to the primary color, the CMYG four-color image signal is converted into an RGB three-color image signal, thereby obtaining the same RGB color image signals as the camera sensor 38. Furthermore, instead of the camera sensor 38, a monochrome sensor without color filters can also be used.

[0043] The CDS / AGC (Correlated Double Sampling / Automatic Gain Control) circuit 46 performs correlation double sampling (CDS) or automatic gain control (AGC) on the analog image signal obtained from the camera sensor 44. The image signal after passing through the CDS / AGC circuit 46 is converted into a digital image signal by the A / D (Analog / Digital) converter 48. The digital image signal after A / D conversion is input to the processor device 16.

[0044] The processor device 16 includes a central control unit 68 that constitutes the processor of the present invention. The central control unit 68 is a hardware resource for executing program commands stored in the memory 69, and drives and controls various parts of the processor device 16 to execute the program commands. Through the drive and control of the central control unit 68 accompanying the execution of program commands, the processor device 16 functions as a camera control unit 50, a DSP (Digital Signal Processor) 52, a noise reduction unit 54, an image processing unit 58, and a display control unit 60.

[0045] The camera control unit 50 includes the aforementioned light source control unit 21 and camera sensor control unit 45. The camera control unit 50 controls the light source unit 20 via the light source control unit 21 to illuminate the object being observed, and controls the camera sensor 44 via the camera sensor control unit 45 to capture images of the object being observed (reading signals from the camera sensor 44). Then, an endoscopic image output from the camera sensor 44 is acquired through this capture.

[0046] The camera control unit 50 performs ordinary imaging by illuminating the object under ordinary light and acquiring an ordinary image. Furthermore, the camera control unit 50 performs special imaging by illuminating the object under special light and acquiring a special image. Special imaging is performed for each type of special light. In this embodiment, endoscopic images, such as ordinary and special images, are color images composed of blue signals (B image signals), green signals (G image signals), and red signals (R image signals) output from the B, G, and R pixels of the camera sensor 44.

[0047] The endoscope images acquired by the camera control unit 50 are sent to the DSP 52. The DSP 52 performs various signal processing on the received endoscope images, including defect correction, offset processing, gain correction, matrix processing, gamma conversion, demosaic processing, and YC conversion.

[0048] In the defect correction process, the signal of the defective pixel of the camera sensor 44 is corrected. In the offset process, the dark current component is removed from the image signal after the defect correction process, and the correct zero level is set. The gain correction process adjusts the signal level of the endoscopic image by multiplying the image signals of various colors after the offset process by a specific gain coefficient. In addition, the endoscopic image is preferably a monochrome image of multiple colors. When a monochrome sensor is used as the camera sensor 44, the monochrome image is captured by the monochrome sensor and output from the monochrome sensor whenever light of a specific color is emitted.

[0049] After gain correction, the image signals of various colors are subjected to matrix processing to improve color reproducibility. Then, gamma conversion is used to adjust the brightness or chroma of the endoscopic image. The matrix-processed endoscopic image is then subjected to demosaicing (also known as isotropic processing or synchronization processing), and signals for the missing colors of each pixel are generated through interpolation. Through demosaicing, all pixels have signals for all RGB colors. The DSP52 performs YC conversion processing on the demosaiced endoscopic image and outputs the brightness signal Y, chromatic aberration signal Cb, and chromatic aberration signal Cr to the noise reduction unit 54.

[0050] The noise reduction unit 54 performs noise reduction processing, such as moving average or median filtering, on the endoscopic image that has undergone de-mosaic processing by the DSP 52. The noise-reduced endoscopic image is then input to the image processing unit 58. In this embodiment, as endoscopic images, both ordinary images (those captured under normal light) and special images (those captured under special light) are input to the image processing unit 58.

[0051] The image processing unit 58 includes a storage device 58a, which stores the input ordinary images and special images. Furthermore, the image processing unit 58 includes a liveness information image generation unit 58b, a recognition processing unit 58c, and a designated image selection unit 58d, which perform various processing operations on the ordinary images and special images.

[0052] The liveness information image generation unit 58b generates a liveness information image representing the liveness information of the observed object using either a special image alone or both a special image and a normal image. The liveness information image is the aforementioned vascular image, oxygen saturation image, chromatic aberration image, etc., and is an image generated by analyzing and processing special images and normal images, and is a type of endoscopic image.

[0053] The recognition processing unit 58c performs recognition processing on ordinary images. This recognition processing includes detection processing to identify regions of interest, such as lesions, and identification processing to determine the type or stage (degree) of the lesions. Furthermore, the identification processing includes processing of the regions of interest and processing of the entire image on which the recognition processing is performed.

[0054] In the identification process, for example, the endoscopic image is segmented into multiple small regions, and image feature quantities are calculated from the segmented endoscopic image. Then, in the detection process, the calculated feature quantities are used to determine whether each small region is a lesion, and a region identified as belonging to the same type is extracted as a lesion. The region including the extracted lesion is then detected as a region of interest. Furthermore, in the discrimination process, regarding the detected region of interest, the type and / or severity (stage) of the lesion are determined based on the feature quantities within the region of interest and the form (location, size, shape, etc.) of the region of interest. The judgments in the above identification process (detection process, discrimination process) are preferably performed using machine learning algorithms such as convolutional neural networks (CNNs) and deep learning.

[0055] Furthermore, recognition processing of ordinary images can be performed using individual ordinary images. Besides ordinary images, special images and / or liveness information images can also be used. An example of recognition processing of ordinary images has been given, but recognition processing of special images and / or liveness information images can also be performed.

[0056] After generating and recognizing the liveness information image, the image processing unit 58 inputs the ordinary image, the liveness information image, and the recognition processing result to the display control unit 60. The display control unit 60 controls the display on the monitor 18, using the input ordinary image, the liveness information image, and the recognition processing result to generate a display screen, which is then displayed on the monitor 18. Furthermore, in the display screen, the liveness information image is overlaid on the ordinary image or displayed side-by-side on the ordinary image. The recognition processing result is displayed, for example, by overlaying a box representing the area of ​​interest onto the ordinary image.

[0057] like Figure 3 As shown, the designated image selection unit 58d operates when a designated operation (operation of the designated switch 12g) is performed on a normal image displayed on the monitor 18 (or an image on which a liveness information image and / or recognition processing result is superimposed). Then, the designated image selection unit 58d performs designated image selection processing to select a designated image.

[0058] like Figure 4As shown, in the designated image selection process, the ordinary image displayed on the monitor 18 when the designated operation is performed, and dozens of frames of ordinary images captured within a specified period (e.g., a few seconds before and after the designated operation) from the capture of the ordinary image, are set as selection target images. Compared with the areas of interest of these selection target images, the image with the highest visibility of the areas of interest is selected as the designated image. In addition, in this embodiment, the higher the contrast, the higher the visibility is considered, and the image with the highest contrast that is determined to be the area of ​​interest is selected as the designated image through image analysis.

[0059] Return to Figure 3 After selecting a specified image through the specified image selection process, the specified image selection unit 58d continues to display the selected specified image on the monitor 18 during the freeze operation if the specified operation is a freeze operation. On the other hand, if the specified operation is a capture operation, the specified image selection unit 58d stores the selected specified image as a still image in the storage device 58a.

[0060] Furthermore, in the above embodiments, the example of higher contrast being considered as higher visibility was explained. However, for example, edge (contour) detection can also be performed, and the more and / or clearer the detected edges, the higher the visibility is considered. Of course, visibility can also be determined based on factors other than contrast and edges. Moreover, as with determining visibility based on only two factors, contrast and edges, multiple factors can also be considered to determine visibility.

[0061] Furthermore, in the above embodiment, an example was described using images captured within a specified period from the start of the specified operation as the selection target when selecting a specified image; however, the present invention is not limited to this. Even images captured within a specified period from the start of the specified operation may contain images with regions of interest such as lesions that are different from the images displayed on the monitor 18 (hereinafter, the displayed images) when the specified operation was performed. Therefore, it is preferable to exclude images with regions of interest such as lesions that are different from the displayed images from the selection target of the specified images.

[0062] Specifically, images whose region of interest (e.g., the center of the region of interest) is separated from the region of interest in the displayed image by a predetermined distance or more are preferably excluded from the selection of designated images. Furthermore, images whose number of regions of interest (detection count) differs from the region of interest in the displayed image, and / or whose area of ​​the region of interest differs from the region of interest in the displayed image by a predetermined size or more, are preferably excluded from the selection of designated images. Additionally, images whose region of interest identification results differ from those of the displayed image are preferably excluded from the selection of designated images.

[0063] Furthermore, if the images captured from the displayed image within a specified period include images in which no region of interest was detected, only images captured within the range (period) from which the region of interest was continuously detected from the displayed image can be selected as the designated images, and images captured outside this range (period) can be excluded from the selection of designated images. Additionally, if the images captured from the displayed image within a specified period include images in which the identification result of the region of interest differs from that of the displayed image, only images captured within the range (period) from which the same identification result was continuously detected from the displayed image can be selected as the designated images, and images captured outside this range (period) can be excluded from the selection of designated images.

[0064] Furthermore, in the above embodiments, examples illustrating the application of the present invention to endoscopic systems (medical images being examples of endoscopic images) have been provided, but the present invention is not limited thereto. The present invention can also be applied to medical systems other than endoscopic systems (medical images can be images other than endoscopic images). Examples of medical systems other than endoscopic systems include ultrasound examination systems that use ultrasound to acquire ultrasound images, and X-ray examination systems that use X-rays to acquire X-ray images.

[0065] In the above embodiments, the hardware structure of the processing units (processing units) that perform various processes, such as the light source control unit 21, the camera sensor control unit 45, the camera control unit 50, the DSP 52, the noise reduction unit 54, the image processing unit 58, the liveness information image generation unit 58b, the recognition processing unit 58c, the designated image selection unit 58d, the display control unit 60, and the central control unit 68, is as shown below. These processors include: general-purpose processors that execute software (programs) and function as various processing units, such as CPUs (Central Processing Units); processors such as FPGAs (Field Programmable Gate Arrays) whose circuit structures can be modified after manufacturing, i.e., programmable logic devices (PLDs); and processors with circuit structures specifically designed for performing various processes, i.e., dedicated circuits.

[0066] A processing unit can be composed of one of these various processors, or it can be composed of a combination of two or more processors of the same or different types (e.g., multiple FPGAs or a combination of a CPU and an FPGA). Furthermore, a single processor can constitute multiple processing units. As examples of a single processor constituting multiple processing units, firstly, in computers such as client and server computers, a processor is composed of a combination of one or more CPUs and software, and this processor functions as multiple processing units. Secondly, in systems-on-a-chip (SoC), a processor is used to implement the functions of the entire system, including multiple processing units, using a single integrated circuit (IC) chip. Thus, as a hardware structure, various processing units can be constructed using one or more of the aforementioned processors.

[0067] More specifically, the hardware structure of these various processors is a circuit that combines circuit elements such as semiconductor components. Furthermore, the hardware structure of the storage section consists of storage devices such as HDDs (hard disk drives) and SSDs (solid state drives).

[0068] Symbol Explanation

[0069] 10-Endoscopic system (medical image processing system), 12-Endoscope, 12a-Insertion section, 12b-Operating section, 12c-Bend section, 12d-Tip section, 12g-Designation switch, 12h-Mode switching switch, 12i-Zoom operating section, 14-Light source device, 16-Processor device, 18-Monitor, 19-UI, 20-Light source section, 21-Light source control section, 23-Optical path coupling section, 25-Light guide, 30a-Illumination optical system, 30b-Imaging optical system. 32-Illumination lens, 42-Objective lens, 43-Zoom lens, 44-Image sensor, 45-Image sensor control unit, 46-CDS / AGC circuit, 48-A / D converter, 50-Image control unit, 52-DSP, 54-Noise reduction unit, 58-Image processing unit, 58a-Memory device, 58b-Liveness information image generation unit, 58c-Recognition processing unit, 58d-Specified image selection unit, 60-Display control unit, 68-Central control unit (processor), 69-Memory.

Claims

1. A medical image processing system, comprising a memory for storing program commands and a processor for executing the program commands. The processor performs the following processing: Multiple medical images generated by continuously capturing images of the observed object are acquired sequentially. The region of interest is detected by performing recognition processing on each of the plurality of medical images, and the plurality of medical images are sequentially output to a monitor for display; and When a designation operation is performed on a medical image displayed on the monitor, the medical image with the highest visibility of the area of ​​interest is selected from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor, as the medical image designated by the designation operation. The following images are excluded from the selection of the specified medical images: images in which the region of interest is set for a lesion that is different from the display image shown on the monitor when the specified operation is performed; images in which the position of the region of interest is separated from the region of interest of the display image by a predetermined distance or more; images in which the number of regions of interest is different from the region of interest of the display image; and images in which the area of ​​the region of interest differs from the region of interest of the display image by a predetermined size or more.

2. A medical image processing system, comprising a memory for storing program commands and a processor for executing the program commands. The processor performs the following processing: Multiple medical images generated by continuously capturing images of the observed object are acquired sequentially. The region of interest is detected by performing recognition processing on each of the plurality of medical images, and the plurality of medical images are sequentially output to a monitor for display; and When a designation operation is performed on a medical image displayed on the monitor, the medical image with the highest visibility of the area of ​​interest is selected from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor, as the medical image designated by the designation operation. If the images captured within a specified period from the display images shown on the monitor during the specified operation include images in which the region of interest was not detected, the images captured during the period in which the region of interest was continuously detected are selected as the objects for performing the specified medical images, and images captured during periods other than the period in which the region of interest was continuously detected are excluded from the objects for performing the specified medical images.

3. A medical image processing system, comprising a memory for storing program commands and a processor for executing the program commands. The processor performs the following processing: Multiple medical images generated by continuously capturing images of the observed object are acquired sequentially. The region of interest is detected by performing recognition processing on each of the plurality of medical images, and the plurality of medical images are sequentially output to a monitor for display; and When a designation operation is performed on a medical image displayed on the monitor, the medical image with the highest visibility of the area of ​​interest is selected from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor, as the medical image designated by the designation operation. The identification process includes an identification process for the region of interest. The following images are excluded from the selection of the specified medical images: images in which the region of interest is defined for a lesion that is different from the display image shown on the monitor when the specified operation is performed; images in which the position of the region of interest is separated from the region of interest in the display image by a predetermined distance or more; images in which the number of regions of interest is different from the region of interest in the display image; images in which the area of ​​the region of interest differs from the region of interest in the display image by a predetermined size or more; and images in which the result of the identification processing of the region of interest is different from the display image.

4. A medical image processing system, comprising a memory for storing program commands and a processor for executing the program commands. The processor performs the following processing: Multiple medical images generated by continuously capturing images of the observed object are acquired sequentially. The region of interest is detected by performing recognition processing on each of the plurality of medical images, and the plurality of medical images are sequentially output to a monitor for display; and When a designation operation is performed on a medical image displayed on the monitor, the medical image with the highest visibility of the area of ​​interest is selected from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor, as the medical image designated by the designation operation. The identification process includes an identification process for the region of interest. If the result of the identification process for the region of interest is different from that of the displayed image, which is captured within a specified period from the image displayed on the monitor during the specified operation, the image captured during the period in which the same identification result is continuously detected is selected as the target for the specified medical image, and the image captured during a period other than the period in which the same identification result is continuously detected is excluded from the target for the specified medical image.

5. The medical image processing system according to any one of claims 1 to 4, wherein, The medical image with the highest contrast was judged to have the highest visibility.

6. The medical image processing system according to any one of claims 1 to 4, wherein, The region including the lesion site is detected through the identification process and designated as the region of interest.

7. The medical image processing system according to claim 5, wherein, The region including the lesion site is detected through the identification process and designated as the region of interest.

8. The medical image processing system according to any one of claims 1 to 4, wherein, The specified operation is a freeze operation that stops the monitor from updating its screen and continues to display a medical image.

9. The medical image processing system according to claim 5, wherein, The specified operation is a freeze operation that stops the monitor from updating its screen and continues to display a medical image.

10. The medical image processing system according to any one of claims 1 to 4, wherein, The specified operation is a capture operation that acquires a medical image displayed on the monitor as a recording image.

11. The medical image processing system according to claim 5, wherein, The specified operation is a capture operation that acquires a medical image displayed on the monitor as a recording image.

12. The medical image processing system according to claim 6, wherein, The specified operation is a capture operation that acquires a medical image displayed on the monitor as a recording image.

13. The medical image processing system according to any one of claims 1 to 4, wherein, In the recognition process, a convolutional neural network is used.

14. The medical image processing system according to claim 5, wherein, In the recognition process, a convolutional neural network is used.

15. The medical image processing system according to claim 6, wherein, In the recognition process, a convolutional neural network is used.

16. The medical image processing system according to claim 8, wherein, In the recognition process, a convolutional neural network is used.

17. The medical image processing system according to any one of claims 1 to 4, wherein, The medical images are images obtained from an endoscope.

18. The medical image processing system according to claim 5, wherein, The medical images are images obtained from an endoscope.

19. The medical image processing system according to claim 6, wherein, The medical images are images obtained from an endoscope.

20. The medical image processing system according to claim 8, wherein, The medical images are images obtained from an endoscope.

21. The medical image processing system according to claim 10, wherein, The medical images are images obtained from an endoscope.

22. A method of operating a medical image processing system, the medical image processing system comprising a memory for storing program commands and a processor for executing the program commands, The processor performs the following processing: Multiple medical images generated by continuously capturing images of the observed object are acquired sequentially. The region of interest is detected by performing recognition processing on each of the plurality of medical images, and the plurality of medical images are sequentially output to a monitor for display; and When a designation operation is performed on a medical image displayed on the monitor, the medical image with the highest visibility of the area of ​​interest is selected from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor, as the medical image designated by the designation operation. The following images are excluded from the selection of the specified medical images: images in which the region of interest is set for a lesion that is different from the display image shown on the monitor when the specified operation is performed; images in which the position of the region of interest is separated from the region of interest of the display image by a predetermined distance or more; images in which the number of regions of interest is different from the region of interest of the display image; and images in which the area of ​​the region of interest differs from the region of interest of the display image by a predetermined size or more.

23. A method of operating a medical image processing system, the medical image processing system comprising a memory for storing program commands and a processor for executing the program commands. The processor performs the following processing: Multiple medical images generated by continuously capturing images of the observed object are acquired sequentially. The region of interest is detected by performing recognition processing on each of the plurality of medical images, and the plurality of medical images are sequentially output to a monitor for display; and When a designation operation is performed on a medical image displayed on the monitor, the medical image with the highest visibility of the area of ​​interest is selected from the medical image displayed on the monitor and medical images captured before and after the medical image displayed on the monitor, as the medical image designated by the designation operation. The identification process includes an identification process for the region of interest. The following images are excluded from the selection of the specified medical images: images in which the region of interest is defined for a lesion that is different from the display image shown on the monitor when the specified operation is performed; images in which the position of the region of interest is separated from the region of interest in the display image by a predetermined distance or more; images in which the number of regions of interest is different from the region of interest in the display image; images in which the area of ​​the region of interest differs from the region of interest in the display image by a predetermined size or more; and images in which the result of the identification processing of the region of interest is different from the display image.

Images