Oral cancer diagnostic device and oral cancer diagnostic program

Abstract

To provide a technique to appropriately conduct a diagnosis of oral cancer.SOLUTION: An oral cancer diagnostic device comprises a control unit that executes: acquiring an image obtained by photographing the inside of the oral cavity of a diagnostic subject; determining a diagnostic target part from the image; and inputting image data of the diagnostic target part determined from the image obtained by photographing the inside of the oral cavity of the diagnostic subject, to a learned diagnostic model subjected to machine learning processing using training data so as to output an estimated invasive depth of oral cancer upon input of the image data of the diagnostic target part, thereby outputting the invasive depth of oral cancer.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an oral cancer diagnosis device and an oral cancer diagnosis program. [Background technology]

[0002] The number of people suffering from oral cancer is increasing year by year, and the cure rate is not high. The most common oral cancer is tongue cancer, and while the 5-year relative survival rate is relatively high in stage I, the 5-year relative survival rate drops significantly in stage IV. However, oral cancer is characterized by leukoplakia and intractable There are many diseases that present similar symptoms, such as ulcers, and diagnosis is difficult unless you are a doctor or dentist specializing in oral surgery or otorhinolaryngology, and there are still few specialists. For this reason, there is a need for the development of a diagnostic system that utilizes AI.

[0003] Patent Document 1 discloses a method for detecting stomatognathic diseases from X-ray images. The detection method includes an input step of inputting an image of the stomatognathic region of a subject obtained by X-ray photography into an analysis unit, a pre-processing step of determining an amplification rate of the image, dividing the image into a pair of left and right regions to be identified based on the determined amplification rate, and amplifying data of the divided images, a sensitivity adjustment step of adjusting sensitivity so as not to overlook the stomatognathic diseases based on the determined amplification rate, and an identification step of learning classifiers corresponding to each of the divided regions to create classifiers, individually identifying each of the divided regions as positive or negative using the classifiers, diagnosing the entire image by combining the results of the identification, and identifying suspected positive areas. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Patent Publication No. 2022-107558 [Patent Document 2] Patent Publication No. 2022-82952 Summary of the Invention [Problem to be solved by the invention]

[0005] If X-ray images are required for diagnosis, as in Patent Document 1, it is necessary to prepare equipment for taking X-ray images, and it takes time and effort to take the X-ray images, which makes it difficult to make a diagnosis.

[0006] Tongue cancer is also a disease that significantly affects functions such as chewing, eating, swallowing, and pronunciation, as well as the cosmetic aspects of the maxillofacial region. Therefore, even if tongue cancer goes into remission and is life-saving, in advanced cases it significantly reduces the quality of life (QOL) in daily life. Although the depth of tumor invasion is an important indicator for predicting such prognosis, there is no established method for evaluating the depth of tumor invasion preoperatively.

[0007] An object of the disclosed technology is to provide a technology for appropriately diagnosing oral cancer. [Means for solving the problem]

[0008] One aspect of the disclosed technology is exemplified by the following configuration of an oral cancer diagnosis device. That is, the oral cancer diagnosis device comprises: Obtaining an image of the oral cavity of a subject; determining a diagnostic target region from the image; A trained diagnostic model that has been subjected to machine learning processing using training data so as to output an estimated depth of invasion of oral cancer when image data of the diagnostic target site is input is inputting image data of the diagnostic target site obtained from the image of the subject's oral cavity, and outputting the depth of invasion of the oral cancer as a diagnostic result; Execute.

[0009] In the oral cancer diagnosis device, the control unit may input image data of the diagnostic target area to the learned diagnostic model, and output the progression of the oral cancer in addition to the depth of invasion as the diagnostic result.

[0010] In the oral cancer diagnosis apparatus, the control unit At least one of the smoking history, drinking history, viral infection history, and leukoplakia history of the subject is obtained as patient information; The depth of invasion of oral cancer may be output by inputting the image data and the patient information into a trained diagnostic model that has been subjected to machine learning processing using teacher data so as to output an estimated depth of invasion of oral cancer when the image data and the patient information are input.

[0011] In the oral cancer diagnosis device, the control unit may output a message prompting an examination when the depth of invasion output from the diagnostic model as the diagnosis result for the subject is equal to or greater than a predetermined value.

[0012] In the oral cancer diagnosis device, the control unit may output a message prompting an examination when the depth of invasion or the degree of progression output from the diagnostic model as the diagnosis result for the subject is equal to or greater than a predetermined value.

[0013] In the oral cancer diagnosis device, the control unit may not output the depth of invasion of the oral cancer when the progression level output from the diagnostic model as the diagnosis result for the subject is less than a predetermined value, and may output the depth of invasion of the oral cancer when the progression level is equal to or greater than a predetermined value.

[0014] In the oral cancer diagnosis apparatus, the control unit determining a feature amount of the diagnostic target region based on at least one of a contour shape, a contour clarity, a contour size, and a concavo-convex state of the diagnostic target region; The depth of invasion of oral cancer may be output by inputting the image data and the feature values ​​into a trained diagnostic model that has been subjected to machine learning processing using teacher data so as to output an estimated depth of invasion of oral cancer when the image data and the feature values ​​are input.

[0015] In the oral cancer diagnosis apparatus, the control unit may acquire images of the inside of the oral cavity of the person to be diagnosed via a network.

[0016] In the oral cancer diagnosis device, the control unit may transmit the diagnosis result to a terminal of the person to be diagnosed or a terminal of a medical institution via a network.

[0017] Moreover, one aspect of the disclosed technology is exemplified by the following oral cancer diagnosis program. That is, the oral cancer diagnosis program is Obtaining an image of the oral cavity of a subject; determining a diagnostic target region from the image; inputting image data of the diagnostic site obtained from the image of the oral cavity of the subject into a trained diagnostic model that has been subjected to machine learning processing using teacher data so as to output an estimated depth of invasion of oral cancer when image data of the diagnostic site is input; and outputting the depth of invasion of the oral cancer. The control unit executes the above.

[0018] Furthermore, one aspect of the disclosed technology is exemplified by a computer-readable recording medium having the oral cancer diagnosis program non-temporarily recorded thereon, and the computer can read and execute the program from the recording medium to provide the function.

[0019] Here, a computer-readable recording medium refers to a recording medium that stores information such as data and programs electrically, magnetically, optically, mechanically, or chemically and can be read by a computer. Among such recording media, those that can be removed from a computer include, for example, flexible disks, magneto-optical disks, CDs (Compact Discs), CD-R / Ws, DVDs (Digital Versatile Disks), Blu-ray Discs (Blu-ray (registered trademark) Discs), DAT, 8mm tapes, flash memories, and other such memories. In addition, there are hard disks and ROMs (read-only memories) as recording media fixed to computers, etc.

[0020] Moreover, one aspect of the disclosed technology is exemplified by the following oral cancer diagnosis method. That is, the oral cancer diagnosis method includes: Obtaining an image of the oral cavity of a subject; determining a diagnostic target region from the image; inputting image data of the diagnostic site obtained from the image of the oral cavity of the subject into a trained diagnostic model that has been subjected to machine learning processing using teacher data so as to output an estimated depth of invasion of oral cancer when image data of the diagnostic site is input; and outputting the depth of invasion of the oral cancer. The control unit executes the above. Effect of the Invention

[0021] The technology disclosed herein provides a technology for appropriately diagnosing oral cancer. [Brief description of the drawings]

[0022] [Figure 1] FIG. 1 is a schematic diagram showing the configuration of an oral cancer diagnosis apparatus (hereinafter also simply referred to as a diagnosis apparatus) according to this embodiment. [Diagram 2] FIG. 2 is a diagram for explaining a process for determining a diagnostic target region from a captured image. [Diagram 3]FIG. 3 is a diagram showing a process in which the control unit of the diagnostic device performs machine learning according to a program. [Figure 4] FIG. 4 is a diagram showing the process executed by the control unit of the diagnostic device in accordance with the oral cancer diagnostic program. [Diagram 5] FIG. 5 is a diagram showing an example of a diagnosis result. [Figure 6] FIG. 6 is a diagram showing an example of a diagnosis result when the invasion depth is not output. [Figure 7] FIG. 7 is a diagram showing the configuration of an oral cancer diagnosis apparatus according to the second embodiment. [Figure 8] FIG. 8 is a diagram showing the configuration of an oral cancer diagnosis apparatus according to the third embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] Hereinafter, an oral cancer diagnosis apparatus according to an embodiment will be described with reference to the drawings. The configuration of the following embodiment is an example, and the oral cancer diagnosis apparatus is not limited to the configuration of the embodiment.

[0024] First Embodiment 《Device configuration》 FIG. 1 is a schematic diagram of an oral cancer diagnosis device (hereinafter also simply referred to as a diagnosis device) 10 according to this embodiment. The diagnosis device 10 inputs an image of the inside of the oral cavity (affected area) of a patient (subject to diagnosis) 20 into a diagnosis model and calculates a diagnosis result of oral cancer. This diagnosis model is based on the results of previous doctors' The diagnosed affected area is photographed, the diagnostic target area is determined from the photographed image, and furthermore, feature amounts related to oral cancer are obtained from the diagnostic target area, and deep learning (machine learning) is performed using the image data of the diagnostic target area, the feature amounts, and the labels determined by the doctor as teacher data. As a result, the diagnostic device 10 of this embodiment can mechanically diagnose the image of the affected area using AI and output the progression and depth of oral cancer.

[0025] Diagnostic Device The diagnostic device 10 is a computer having a control unit 12, a memory 13, an input / output IF (interface) 14, a communication IF 15, and an imaging device 16, all of which are interconnected by a connection bus 11. The control unit 12 processes input information and outputs the processing results to control the entire device. The control unit 12 is also called a CPU (Central Processing Unit) or an MPU (Micro-processing unit). The control unit 12 is not limited to a single processor, and may have a multiprocessor configuration. Also, the control unit 12 may have a multicore configuration having multiple cores in a single chip connected by a single socket.

[0026] The memory 13 includes a main storage device and an auxiliary storage device. The main storage device is used as a working area for the control unit 12, a storage area for temporarily storing information processed by the control unit 12, and a buffer area for communication data. The main storage device includes, for example, a random access memory (RAM), a read only memory (ROM), and a flash memory. The auxiliary storage device is a storage medium for storing programs executed by the control unit 12, data used in information processing, operation setting information, data used as a diagnostic model, and the like. The auxiliary storage device includes, for example, a hard-disk drive (HDD), a solid state drive (SSD), an erasable programmable ROM (EPROM), a flash memory, a USB memory, a memory card, and the like. In addition, the auxiliary storage device includes, for example, a hard disk drive (HDD), a solid state drive (SSD), an erasable programmable ROM (EPROM), a flash memory, a USB memory, a memory card, and the like. The device may store images (photographed images) to be diagnosed and the results of diagnosis.

[0027] The input / output IF 14 is an interface that inputs and outputs data between devices connected to the diagnostic device 10. The input / output IF 14 inputs and outputs data between devices such as a disk drive that reads data from a storage medium such as a CD or DVD, an operation unit, a display device, and the imaging device 16. The operation unit is an input unit such as a mouse, keyboard, or touch panel that inputs information to the diagnostic device 10 by an operator's operation. The display device is an output unit that displays and outputs information such as processing results to the operator.

[0028] The communication IF 15 is an interface (communication module) that communicates with other devices via a communication line (network), and is also called a CCU (Communication Control Unit). For example, the diagnostic device 10 may be composed of multiple computers installed in the home of the patient 20 and in a medical institution, and each computer may have the communication IF 15 as shown in Fig. 1 and transmit and receive images of the oral cavity and processing results to and from a computer installed in another location.

[0029] The imaging device 16 photographs the affected area, acquires the photographed image (image data), and inputs it to the control unit 12 via the input / output IF 14. The imaging device 16 is a so-called camera, but may be a three-dimensional sensor or a ToF (Time of Flight) sensor as long as it can acquire image data. The imaging device 16 may also be a stereo camera capable of acquiring three-dimensional information by photographing the affected area with two cameras spaced apart in the horizontal direction.

[0030] The photographing device 16 may include a lighting device and may be configured to photograph the affected area while illuminating it with illumination light of a specified wavelength. Alternatively, the photographing device 16 may include a lighting device and photograph the affected area while the positional relationship between the lighting device and the imaging elements such as the photographing lens and the imaging element is specified, thereby photographing the affected area while specifying the manner in which shading is performed. The photographing device 16 may be equipped with a sensor that specifies the attitude of the photographing device 16 by detecting gravitational acceleration, etc., and may output information (attitude information) indicating the attitude of the camera at the time of photographing. In other words, the photographing position can be determined from the attitude information. It may be possible to distinguish whether the image was taken facing upward or downward.

[0031] The diagnostic device 10 may include a plurality of each of the above-mentioned components 11 to 16, or some of the components may be omitted. For example, the diagnostic device 10 may not include the imaging device 16, and may acquire an image of the patient 20 captured by an external imaging device via the input / output IF 14 and the communication IF 15.

[0032] In the diagnostic device 10 of this embodiment, the control unit 12 executes an application program, whereby the control unit 12 functions as each processing unit, such as a diagnostic region determination unit 111, a preprocessing unit 112, a feature amount calculation unit 113, a patient information acquisition unit 114, a diagnosis unit 115, a result output unit 116, and a machine learning unit 117. That is, the control unit 12 can be used as each processing unit depending on the software being executed. However, it is also possible that some or all of the above processing units are implemented using dedicated LSIs (large scale integration) such as DSPs (Digital Signal Processors), ASICs (Application Specific Integrated Circuits), and FPGAs (Field-Programmable Gate Arrays), logic circuits, etc. The processing units may be formed of hardware such as digital circuits, analog circuits, or other digital circuits. At least a part of each of the processing units may include an analog circuit. The control unit 12 may be configured such that one processor functions as multiple processing units, or may be configured to include multiple processors that function as one processing unit.

[0033] The diagnostic area determination unit 111 acquires an image of the affected area photographed by the photographing device 16 from the photographing device 16, or acquires an image of the affected area photographed by another device via the input / output IF 14 or the communication IF 15, and determines the part of the photographed image to be used for diagnosis (the area to be diagnosed).

[0034] For example, Fig. 2 is a diagram for explaining the process of determining a diagnostic target part from a photographed image. As shown in Fig. 2, the diagnostic target part determination unit 111 performs edge detection on the acquired photographed image 51, identifies an area 52 surrounded by this edge, and determines a diagnostic target part 53 so as to include at least the area 52. Here, the diagnostic target part determination unit 111 may determine only the area 52 surrounded by the edge as the diagnostic target part 53, or may determine a rectangular area including the area 52 as the diagnostic target part 53.

[0035] The diagnostic region determining unit 111 may determine whether the size, planar shape, unevenness (three-dimensional shape), color, etc. of the region 52 surrounded by the edge satisfy a predetermined condition, and determine the region that satisfies the condition as the diagnostic region. When a plurality of regions 52 surrounded by the edge are detected, the diagnostic region determining unit 111 may determine whether these regions 52 satisfy a plurality of conditions related to the size, planar shape, unevenness (three-dimensional shape), color, etc., and narrow down the region to be determined as the diagnostic region according to the number of conditions that are satisfied. In addition, the region to be determined as the diagnostic region may be narrowed down by identifying other regions present in the oral cavity, such as the patient's teeth and uvula, by pattern matching, and excluding these.

[0036] Furthermore, the diagnostic region determination unit 111 may obtain the size of the diseased region based on the imaging conditions such as the imaging magnification and the distance to the diseased region when the diseased region is imaged by the imaging device 16, and the size (e.g., the number of pixels) of the area 52 captured in the captured image, and may set this as a parameter of the size of the diagnostic region. Note that when the diseased region is imaged by the imaging device 16, the image may be captured so that the diseased region is imaged at a specified size, and thus input of the parameter relating to the size may be omitted.

[0037] The preprocessing unit 112 performs preprocessing on the diagnostic target region (image data) determined by the diagnostic region determining unit 111 so that the size, color, brightness, contrast, etc. are in a predetermined state. For example, if the color, contrast, etc. of the diagnostic target region varies depending on the imaging conditions, etc., the color and shape of the affected area cannot be evaluated correctly, so the condition of the diagnostic target region is normalized.

[0038] The feature amount calculation unit 113 calculates the feature amount of the diagnosis target part based on at least one of the contour shape, contour clarity, contour size, and unevenness of the diagnosis target part normalized by the preprocessing unit 112. When the feature amount is not used in the diagnosis process, the feature amount calculation unit 113 may be omitted.

[0039] For example, a baseline may be set based on the shape of the periphery of the diagnostic region, and the height of the central area of ​​the diagnostic region relative to this baseline may be quantified and used as a parameter for the uneven state.

[0040] The patient information acquiring unit 114 acquires patient information indicating the age, lifestyle, and the like of the patient. For example, the patient information acquiring unit 114 displays an input form (input screen) of the patient information on the display device to prompt the operator to input, and acquires the patient information input via the operation unit. The patient information acquiring unit 114 may also be connected to an electronic medical record system or the like and acquire the patient information from a server in which the patient information is registered. Here, the patient information is, for example, the number of teeth (the number of teeth that are not dentures), the history of denture use, the smoking history, the drinking history, the age, the history of viral infection, the history of leukoplakia, the history of xerostomia, inflammation, and the like. The history of denture use is, for example, information such as the type, shape, start date of use, and the time of wearing of the dentures per day. The smoking history is, for example, information such as the number of cigarettes smoked per day (smoking amount) and the period during which the smoking habit was continued (smoking period). Furthermore, drinking history is information such as the amount of alcohol consumed per day, the number of days per week that alcohol is consumed, the main type of alcohol consumed, the alcohol content of the main alcohol consumed, and the period during which the drinking habit was continued (drinking period). Viral infection history is information such as the type of virus that infected a person in the past, the details of symptoms, the severity of symptoms, the date of onset, and the date of cure or remission. Leukoplakia history is information such as the severity of symptoms, the size of leukoplakia, the date of onset, and the date of cure or remission. Xerostomia history is information such as the severity of symptoms, the date of onset, and the date of cure or remission. Inflammation is information such as white blood cell count, red blood cell count, hemoglobin amount, albumin amount, CRP value, etc. For items for which a unique quantitative value can be calculated, such as the number of teeth, age, and amount of alcohol consumed per day, these values ​​are input to diagnostic device 10. For items consisting of multiple elements, such as viral infection history, smoking history, and alcohol consumption history, multiple axes are set, for example, in an N-dimensional space, and data in which the values ​​of each element are recorded for each axis are input to diagnostic device 10. For example, in the case of a viral infection history, multiple axes (e.g., three axes each) are set for each infectious disease, and Bacterium A infection first axis: presence or absence of infection, Bacterium A infection second axis: presence or absence of infection, The second axis: age at infection (period from infection to recovery (N years M months to N years M months, or D days)), and the third axis: severity (mild, moderate, severe), etc. are recorded. In the case of smoking history, the number of cigarettes smoked per day (smoking amount), the period during which the smoking habit was continued (smoking period), and the type of smoking implement (cigarettes, electronic cigarettes (non-combustible cigarettes), cigars, chewing tobacco, etc.) may be recorded on each axis. In the case of leukoplakia history, for example, the severity of symptoms, the size of leukoplakia, the date of onset, the date of complete recovery or remission, etc. may be recorded on each axis. In addition, the above elements such as xerostomia history, drinking history, and inflammation may also be recorded on each axis. In addition, when patient information is not used in the diagnosis process, the patient information acquisition unit 114 may be omitted.

[0041] The diagnosis unit 115 calculates the progression and depth of invasion of oral cancer by inputting image data of the diagnosis target area determined by the diagnosis area determination unit 111 and the preprocessing unit 112 for a moving image captured of the patient 20 into the diagnosis model. The diagnosis unit 115 of this embodiment calculates the progression and depth of invasion of oral cancer by inputting, in addition to image data, the feature amount determined by the feature amount calculation unit 113 and the patient information acquired by the patient information acquisition unit 114 into the diagnosis model. Note that the diagnosis model of this embodiment is configured to output the progression and depth of invasion of oral cancer, but is not limited to this, and may output only the invasion depth.

[0042] The machine learning unit 117 generates a diagnostic model by performing machine learning based on the learning data. Note that a specific learning method will be described later.

[0043] The progression of oral cancer is, for example, classified into the following five stages. A progression value of 1 (stage 1) indicates a healthy state in which oral cancer has not progressed. A progression value of 5 (stage 5) indicates the most advanced state of oral cancer. In other words, the higher the progression value, the more advanced the oral cancer is. Stage 5: Oral cancer is present (there are atypical cells that are definitively malignant) Stage 4: Oral cancer is present or there is a precancerous lesion (high possibility of malignant cells, or carcinoma in situ) Stage 3: Condition requiring observation (mild to moderate dysplasia) Stage 2: No need to see a specialist (mouth ulcers and other minor cases that can be treated at a clinic without any problems (atypical cells are present but not malignant)) Stage 1: No findings (normal cells) The degree of progress is not limited to this, and may be determined in multiple stages based on other criteria.

[0044] The depth of invasion of oral cancer is, for example, classified into the following three stages: The greater the value of the depth of invasion, the deeper the oral cancer is present. Stage 3: Depth of invasion exceeds 10 mm Stage 2: Depth of invasion: greater than 5mm, less than 10mm Stage 1: Penetration depth 5mm or less The invasion depth is not limited to this, and may be determined in multiple stages based on other criteria.

[0045] The result output unit 116 outputs the degree of progression and the depth of invasion of oral cancer calculated by the diagnosis unit 115 by displaying them on a display unit, etc. Note that the output of the degree of symptoms of oral cancer is not limited to displaying them, and may be printing, transmitting them to another device, or recording them on a recording medium.

[0046] Machine learning processing 3 is a diagram showing a process in which the control unit 12 (machine learning unit 117) of the diagnostic device 10 performs machine learning according to a program. When an instruction to start machine learning is received, the control unit 12 executes the process in FIG.

[0047] In step S10, the control unit 12 acquires learning data from an auxiliary storage device or a storage medium connected to the input / output IF 14. The learning data (teacher data) includes, for example, images of the oral cavity of the patient 20 captured in advance for machine learning, and teacher labels in which a doctor or the like has evaluated the progression and depth of oral cancer of the patient 20. The learning data may further include patient information and feature amounts. In addition, when the progression is not output as a diagnosis result, the teacher labels of the progression may be omitted.

[0048] In step S20, the control unit 12 determines a diagnostic target region from the image acquired in step S10. Note that, if the image acquired in step S10 is an image of a diagnostic target region, step S20 may be omitted.

[0049] In step S30, the control unit 12 performs pre-processing for normalizing the size, color, brightness, contrast, etc. of the diagnosis target region obtained in step S20.

[0050] In step S40, the control unit 12 obtains the feature amounts of the diagnosis target part. In this embodiment, the feature amounts obtained are the shape of the contour, the clarity of the contour, the size of the contour, and the state of unevenness of the contour in the diagnosis target part.

[0051] In step S50, the control unit 12 performs machine learning using the feature amounts obtained in step S40, the image data of the diagnostic target area, and the teacher labels obtained in step S10 as learning data, and generates a diagnostic model (trained diagnostic model).

[0052] In step S60, the control unit 12 determines whether or not learning has been completed for all learning data, and if the determination is negative, the process returns to step S10, whereas if the determination is positive, the process of FIG. 3 ends.

[0053] Oral Cancer Diagnosis and Treatment 4 is a diagram showing a process executed according to an oral cancer diagnosis program by the control unit 12 of the diagnostic device 10. When an instruction to start diagnosis is given, the control unit 12 executes the process of FIG.

[0054] In step S110, the control unit 12 acquires an image of the inside of the oral cavity of the patient 20 from the imaging device 16, an auxiliary storage device, or a storage medium connected to the input / output IF .

[0055] In step S120, the control unit 12 determines a diagnosis target region from the image acquired in step S110. The diagnosis target region determined here is the same as that determined in step S20 of the machine learning process shown in FIG.

[0056] In step S130, the control unit 12 performs pre-processing for normalizing the size, color, brightness, contrast, and the like of the diagnosis target region obtained in step S120.

[0057] In step S140, the control unit 12 obtains the feature amount of the diagnosis target part. In this embodiment, the feature amount is obtained as the shape of the outline, the sharpness of the outline, the size of the outline, and the state of unevenness in the diagnosis target part. Note that, if the feature amount is not used for diagnosis, step S140 may be omitted.

[0058] In step S150, the control unit 12 inputs the feature values ​​calculated in step S140 and the image data of the diagnostic target area into the diagnostic model, and outputs the stage and depth of invasion of oral cancer for the patient 20. When the feature values ​​are not used in the diagnosis, the control unit 12 may omit inputting the feature values ​​and output the stage and depth of invasion of oral cancer based on the image data.

[0059] In step S160, the control unit 12 judges whether the progression or depth of invasion calculated in step S150 is equal to or greater than a predetermined value. If the judgment in step S160 is positive, the control unit 12 proceeds to step S170 and generates a message (examination message) to prompt the patient 20 to undergo an examination. The predetermined value is set according to the stage at which a detailed examination is recommended, such as stage 2 or higher, or stage 3 or higher. Also, the predetermined value for judging the depth of invasion and the predetermined value for judging the progression may be set to different values, such as stage 2 or higher for the depth of invasion and stage 3 or higher for the progression. When the generation of this examination message is completed and when the judgment in step S160 is negative, the control unit 12 proceeds to step S180.

[0060] In step S180, the control unit 12 determines whether the degree of progression is less than a predetermined value. If the determination in step S180 is positive, the control unit 12 proceeds to step S190 and cancels the output of the invasion depth. That is, the control unit 12 sets the value of the invasion depth to a null value and prevents the output of the invasion depth to the patient 20. Thereby, for example, when the degree of progression is stage 1 (no findings), the output of the invasion depth may not be performed. When step S190 is completed, or when the determination in step S180 is negative, the control unit 12 proceeds to step S200.

[0061] In step S200, the control unit 12 calculates the degree of progression of oral cancer and The diagnostic device 10 outputs the degree of invasion and the depth of invasion, and the examination message generated in step S170 as the diagnostic result. FIG. 5 is a diagram showing an example of the diagnostic result. In the example of FIG. 5, the diagnostic result 61 includes an image 62 of the affected area, a progression degree 63, a depth of invasion 64, and an examination message 65. The diagnostic result 61 is presented (display output) to the patient by being displayed on a display device. FIG. 6 is a diagram showing an example of the diagnostic result when the depth of invasion is not output. In the example of FIG. 6, the value of the depth of invasion is canceled in step S190, and the diagnostic result 61 does not include the value of the depth of invasion 64. In the example of FIG. 6, it is determined (negative determination) that the progression degree or the depth of invasion is not equal to or greater than a predetermined value in step S160, and the examination message is not output. Note that the display of the diagnostic result 61 is not limited to being displayed on a display device provided in the diagnostic device 10, but may be transmitted from the diagnostic device 10 to a smartphone or the like (user terminal) of the patient and displayed on a display device of the user terminal. Moreover, the output of the diagnosis result 61 is not limited to being displayed, but may be output as a voice message, printed out, recorded on a recording medium, transmitted to another device, or the like.

[0062] In this way, the diagnostic device 10 of this embodiment determines the diagnostic target site from an image of the patient 20, obtains the feature amount of this diagnostic target site, and outputs the progression and invasion depth of oral cancer by AI based on this feature amount. As a result, the diagnostic device 10 of this embodiment can provide a technology for appropriately diagnosing oral cancer. For example, even a doctor who is not a specialist or a doctor with little experience can appropriately diagnose oral cancer based on the progression and invasion depth output by the diagnostic device 10, and can take appropriate measures such as referring a specialist as necessary. In addition, the diagnostic device 10 of this embodiment can output the progression based on past data learned by the AI ​​model even if the affected area does not clearly show the characteristics of oral cancer, so that it can contribute to early detection of oral cancer. Furthermore, since the diagnostic device 10 of this embodiment outputs an estimated invasion depth, a doctor can appropriately predict the prognosis, etc.

[0063] Second Embodiment In the first embodiment described above, an example of a standalone oral cancer diagnosis device 10 was shown, but the oral cancer diagnosis device 10A of this embodiment is an online diagnosis system that performs diagnosis via a network. The oral cancer diagnosis device 10A of this embodiment is different from the above embodiment in the configuration for communicating data via a network, but the other configurations are the same. Therefore, the same elements as those in the first embodiment are given the same reference numerals, and repeated explanations will be omitted.

[0064] 7 is a diagram showing the configuration of an oral cancer diagnosis apparatus 10A according to the second embodiment. The oral cancer diagnosis apparatus 10A of this embodiment is connected to a user terminal 21 and a doctor terminal 66 via a network 40 such as the Internet.

[0065] The user terminal 21 is an information processing device such as a smartphone or a tablet PC (personal computer) used by a patient (user) 20. The user terminal 21 is the oral cancer diagnostic device shown in FIG. Like the device 10, it has a control unit, memory, an input / output IF (interface), a communication IF, and an imaging device, and has the function of photographing the patient's oral cavity to generate an image, and the function of transmitting the image to the oral cancer diagnosis device 10A via the network 40.

[0066] The doctor terminal 66 is an information processing device used by the doctor 60 and hospital staff such as nurses and medical technicians. The doctor terminal 66 may be one terminal of a so-called hospital information system (HIS). Like the oral cancer diagnosis device 10 in Fig. 1, the doctor terminal 66 has a control unit, a memory, an input / output IF (interface), a communication IF, and an imaging device, and has a function of receiving the diagnosis result of the oral cancer diagnosis device 10A via the network 40 and presenting it to the doctor 60, etc.

[0067] The oral cancer diagnosis apparatus 10A is a server that performs so-called cloud computing. Similar to the oral cancer diagnosis device 10 of the embodiment, the oral cancer diagnosis device 10A has a control unit 12, a memory 13, an input / output IF 14, a communication IF 15, and an imaging device 16. Similarly to the first embodiment, the control unit 12 of the oral cancer diagnosis device 10A functions as each processing unit such as a diagnostic site determination unit 111, a preprocessing unit 112, a feature calculation unit 113, a patient information acquisition unit 114, a diagnosis unit 115, a result output unit 116, and a machine learning unit 117. Note that the oral cancer diagnosis device 10A of the present embodiment may omit the imaging device 16 by acquiring captured images from a user terminal.

[0068] The diagnostic region determining unit 111 of this embodiment uses, for example, the communication IF 15 to connect to the user terminal 21 via the network 40, acquires an image of the affected area of ​​the patient 20 captured by an imaging device from the user terminal 21, and determines a portion of the captured image to be used for diagnosis (diagnosis target region). Note that the functions of the preprocessing unit 112, feature amount calculating unit 113, patient information acquiring unit 114, diagnosing unit 115, and the like that perform diagnosis based on the captured image are similar to those of the above-mentioned embodiment.

[0069] The result output unit 116 uses, for example, the communication IF 15 to connect to the doctor terminal 66 via the network 40, and outputs the progression and depth of the oral cancer calculated by the diagnosis unit 115 by transmitting them to the doctor terminal 66. Note that the result output unit 116 is not limited to directly connecting to the doctor terminal 66 to transmit the diagnosis results, but may present the diagnosis results to the doctor 60, etc., by transmitting the diagnosis results to an HIS such as an electronic medical record system or an ordering system to which the doctor terminal 66 can be connected.

[0070] The doctor 60 who has been presented with the diagnosis results understands the condition of the affected area based on the diagnosis results, and if necessary, makes an appointment for an examination to make a definitive diagnosis, or an appointment for surgery to remove part of the diseased area.

[0071] In this way, according to the oral cancer diagnosis device 10A of the present embodiment, the patient can obtain the diagnosis result of the oral cancer diagnosis device 10A without going to a hospital where a doctor 60 is present. Therefore, for example, even if the patient does not think that it is necessary to go to a hospital, the patient can obtain the diagnosis result by simply taking a picture of the affected area with his / her own smartphone or the like and sending it to the oral cancer diagnosis device 10A, which can lead to early detection of oral cancer. In addition, a patient who has been treated for oral cancer can periodically take a picture of the affected area and send the taken image from the user terminal 21 to the oral cancer diagnosis device 10A, and if there is no finding as a result of the diagnosis, nothing is done, and if the progression or depth of the oral cancer is equal to or greater than a predetermined value, a medical examination is scheduled, thereby allowing follow-up after treatment. In this case, the patient does not need to go to a hospital for each diagnosis by the oral cancer diagnosis device 10A, which reduces the burden on the patient.

[0072] <Third embodiment> The oral cancer diagnosis device 10B of this embodiment is different from the oral cancer diagnosis device 10A of the second embodiment in that the diagnosis result is returned to the user terminal 21 instead of the doctor terminal 66. Since the other configurations are the same, the same elements as those in the second embodiment are denoted by the same reference numerals, and a repeated description will be omitted.

[0073] FIG. 8 is a diagram showing the configuration of an oral cancer diagnosis device 10B according to the third embodiment. In the oral cancer diagnosis device 10B of this embodiment, the diagnosis site determination unit 111 acquires a captured image of the affected area from the user terminal 21, and the result output unit 116 returns the progression and invasion depth of oral cancer calculated by the diagnosis unit 115 to the user terminal 21. Note that the result output unit 116 may not output a message encouraging a patient to visit a clinic if the progression or invasion depth is less than a predetermined value, such as no findings, as a result of diagnosis, and may output a message encouraging a patient to visit a clinic if the progression or invasion depth is equal to or greater than a predetermined value. When outputting a message encouraging a patient to visit a clinic, the oral cancer diagnosis device 10B may, for example, present a selection of medical institutions and make an appointment for a medical examination at the medical institution selected by the patient.

[0074] The patient (user) 20 who is presented with the diagnosis result can understand the condition of the affected area based on the diagnosis result. and, if necessary, make appointments for medical examinations and tests.

[0075] In this way, according to the oral cancer diagnosis device 10B of the present embodiment, the patient can obtain the diagnosis result of the oral cancer diagnosis device 10A without going through a medical institution. Therefore, the patient can easily receive a diagnosis and obtain an appropriate diagnosis result. For example, it is possible to prevent a patient from self-diagnosing oral cancer as stomatitis and delaying a medical examination.

[0076] "others" The above embodiment is merely an example and is not limited to the configurations illustrated in the specification and drawings. For example, the diagnostic device 10 described above has a function of performing machine learning and a function of performing diagnosis, but the device performing machine learning and the device performing diagnosis may be configured separately. In addition, the diagnostic device 10 may be a server that obtains an image via a network and returns the result of diagnosis based on the image. [Explanation of symbols]

[0077] 10: Diagnostic equipment 11: Connection bus 12: Control unit 13: Memory 14: Input / Output IF 15: Communication IF 16: Photographing equipment 20:Patient 51: Photographed image 53: Diagnostic area 111: Diagnostic site determination unit 112: Pre-processing section 113: Feature calculation unit 114: Patient information acquisition department 115: Diagnostics Department 116: Result output section 117: Machine Learning Department

Claims

1. Obtaining an image of the oral cavity of a subject; determining a diagnostic target region from the image; inputting image data of the diagnostic site obtained from the image of the oral cavity of the subject into a trained diagnostic model that has been subjected to machine learning processing using teacher data so as to output an estimated depth of invasion of oral cancer when image data of the diagnostic site is input; and outputting the depth of invasion of the oral cancer as a diagnostic result; An oral cancer diagnosis device comprising a control unit that executes the above.

2. The oral cancer diagnosis device of claim 1 , wherein the control unit inputs image data of the diagnostic target area into the learned diagnostic model, and outputs the progression of the oral cancer in addition to the depth of invasion as the diagnostic result.

3. The control unit: At least one of the smoking history, drinking history, viral infection history, and leukoplakia history of the subject is obtained as patient information; inputting the image data and the patient information into a trained diagnostic model that has been subjected to machine learning processing using training data so as to output an estimated depth of invasion of oral cancer when the image data and the patient information are input; and outputting the depth of invasion of the oral cancer. The oral cancer diagnostic device according to claim 1 .

4. The oral cancer diagnosis device according to claim 1 , wherein the control unit outputs a message prompting an examination when the depth of invasion output from the diagnostic model as the diagnosis result for the subject is equal to or greater than a predetermined value.

5. The oral cancer diagnosis device of claim 2 , wherein the control unit outputs a message prompting an examination when the depth of invasion or the degree of progression output from the diagnostic model as the diagnosis result for the subject is equal to or greater than a predetermined value.

6. The oral cancer diagnosis device of claim 2, wherein the control unit does not output the depth of invasion of the oral cancer when the progression level output from the diagnostic model as the diagnosis result for the subject is less than a predetermined value, and outputs the depth of invasion of the oral cancer when the progression level is equal to or greater than a predetermined value.

7. The control unit: determining a feature amount of the diagnostic target region based on at least one of a contour shape, a contour clarity, a contour size, and a concavo-convex state of the diagnostic target region; outputting the depth of invasion of oral cancer by inputting the image data and the feature amount into a trained diagnostic model that has been subjected to machine learning processing using training data so as to output an estimated depth of invasion of oral cancer when the image data and the feature amount are input; The oral cancer diagnostic device according to any one of claims 1 to 6.

8. 7. The oral cancer diagnostic device according to claim 1, wherein the control unit acquires images of the oral cavity of the subject via a network.

9. The oral cancer diagnostic device according to any one of claims 1 to 6, wherein the control unit transmits the diagnostic result to a terminal of the person to be diagnosed or a terminal of a medical institution via a network.

10. Obtaining an image of the oral cavity of a subject; determining a diagnostic target region from the image; inputting image data of the diagnostic site obtained from the image of the oral cavity of the subject into a trained diagnostic model that has been subjected to machine learning processing using teacher data so as to output an estimated depth of invasion of oral cancer when image data of the diagnostic site is input; and outputting the depth of invasion of the oral cancer. An oral cancer diagnosis program for causing a control unit to execute the above.

Images