System and method for indicating part of colonoscopic examination

Abstract

A method for indicating a part of colonoscopic examination according to the present invention comprises the steps in which: a model loading / condition setting unit loads a colonoscopy image analysis model and sets analysis conditions of the analysis model; a control unit initializes an analysis screen and displays a picture of a normal large intestine; an image preprocessing unit preprocesses a colonoscopy image received via an image reception unit so that subsequent image analysis may be smoothly performed; an image analysis unit analyzes the preprocessed colonoscopy image by using an AI-based image analysis model; the image analysis unit detects and indicates at least one of an examination part, a diverticulum part, or a lesion part from the colonoscopy image on the basis of the analyzed result; and the control unit provides the result analyzed by the image analysis unit.

Description

Colonoscopy Examination Site Indication System and Method

[0001] The present invention relates to a system and method for indicating a colonoscopy examination site, and more specifically, to a system and method for indicating a major examination site by applying image recognition technology during the colonoscopy examination process, wherein the system notifies the examiner of a situation such as a diverticulum protruding from the colon wall or indicates on the examination screen a state different from a normal colon.

[0002] The national research and development projects that supported this invention are as follows.

[0003] Project unique number not assigned

[0004] Assignment Number P0026379

[0005] Ministry of Trade, Industry and Energy

[0006] Project Management Agency Name: Korea Institute for Industrial Technology Promotion

[0007] Research Project Name (24-25) Scale-up Technology Commercialization Program R&D Support (Phase 2)

[0008] Research Project Title: Development of Advanced Gastrointestinal Endoscopy Image Analysis System and Diagnosis Reporting AI System

[0009] Project Executing Organization Name: Waysen (Lead) / Junseong Patent (Joint)

[0010] Research Period: 2024.01.01 ~ 2025.12.31

[0011] Project unique number not assigned

[0012] Project Number RS-2024-00510314

[0013] Ministry of SMEs and Startups

[0014] Project Management Agency Name: Korea Technology Information Promotion Agency for SMEs

[0015] Research Project Name: Startup Growth Technology Development (R&D)

[0016] Research Project Title: Development of AI-based Technology for Improving Colonoscopy Quality - Technology for Appendix Detection, Endoscopy Speed, and Examination Time Measurement

[0017] Project Executing Organization Name: Waysen Co., Ltd.

[0018] Research Period: 2024.10.01 ~ 2025.09.30

[0019] Today, the incidence of colorectal cancer is increasing even more rapidly due to Westernized dietary habits and advancements in diagnostic technology. It is important to note that 80 to 90% of colorectal cancers originate from polyps (adenomas), which are small growths in the colon. If these polyps are detected and removed early through colonoscopy, the mortality rate from colorectal cancer can be significantly reduced.

[0020] The purpose of colorectal cancer screening is to reduce cancer-related mortality by detecting the disease early. According to research results to date, the effectiveness of screening in reducing cancer mortality varies depending on the screening method. Cancer screening methods must offer high sensitivity and specificity, be free from risks or complications, and be affordable. Currently proposed screening methods include the fecal occult blood test (FEBC), sigmoidoscopy, colonoscopy, and double-contrast barium enema. Large-scale randomized clinical trials conducted in Western Europe have reported that the FEBC reduces colorectal cancer mortality by 15–33%. While the FEBC offers advantages such as the absence of complications, low cost, and a relatively simple procedure, issues such as low sensitivity and positive predictive value with a single test, as well as the need for additional testing due to a high false positive rate, have been pointed out.

[0021] Therefore, although screening using colonoscopy is recommended recently, it is being applied only restrictively in national screenings targeting the general public (such as additional testing for individuals with abnormal fecal occult blood test results) due to the relatively high cost, rare but serious complications (such as colon perforation), pain and discomfort for the examinee caused by pre-treatment, and a shortage of skilled endoscopists.

[0022] Meanwhile, Korean Published Patent Application No. 10-2022-0140924 discloses "deep learning-based colonoscopy image analysis method and image analysis system using the same." The deep learning-based colonoscopy image analysis system according to the above is characterized by comprising: an endoscope computer that receives an image captured by a colonoscopy device; a server that acquires the image transmitted to the endoscope computer via a hooking method through an application downloaded to the endoscope computer, and is equipped with a diagnostic algorithm that performs correction of the acquired image and examination based on the corrected image based on deep learning, and transmits the diagnostic result derived through the corrected image and the diagnostic algorithm to the endoscope computer; and a display device that receives the corrected image and the diagnostic result from the endoscope computer and outputs them.

[0023] In the case of the aforementioned patent document, although it has the advantage of being able to operate without the manufacturer's API (Application Programming Interface) by loading and processing images using a window hooking method, and improving diagnostic reliability by correcting images caused by light reflection and accurately measuring the size of polyps, it does not provide a separate means to notify the examiner of conditions such as diverticula during the colonoscopy process or to indicate (display) them on the examination screen differently from a normal colon, thereby carrying the risk that the examiner may cause complications such as perforation.

[0024] The present invention was created by comprehensively considering the above-mentioned matters, and aims to provide a colonoscopy examination site indication system and method that apply image recognition technology to the colonoscopy examination process to indicate key examination sites, thereby notifying the examiner of situations such as diverticula protruding from the colon wall or indicating (displaying) them on the examination screen differently from a normal colon, allowing the examiner to proceed with the examination while paying attention to complications such as perforation, assisting in setting the probe movement path, and enabling the examiner to perform a more thorough colonoscopy.

[0025] In addition, another objective of the present invention is to provide a colonoscopy examination site indication system and method that allows verification of the areas examined to date by indicating major colon examination sites, serves as an indicator of colonoscopy examination quality by indicating whether the cecum has been examined, and records the presence of diverticula, thereby enabling the examiner to proceed with the next colonoscopy examination while being aware of the risk of perforation in advance.

[0026] To achieve the above objective, a colonoscopy examination site indication system according to one embodiment of the present invention,

[0027] A model loading / condition setting unit that loads a colonoscopy image analysis model and sets the analysis conditions of the analysis model;

[0028] An image receiver that receives colonoscopy image frames;

[0029] An image preprocessing unit that preprocesses a colonoscopy image received through the image receiving unit to facilitate subsequent image analysis;

[0030] An image analysis unit that analyzes a colonoscopy image preprocessed by the image preprocessing unit using an AI (artificial intelligence)-based image analysis model, and detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image based on the analyzed result; and

[0031] The invention is characterized by including a control unit that controls the status check and operation of the above-mentioned model loading / condition setting unit, image receiving unit, image preprocessing unit, and image analysis unit, and when the loading of the colonoscopy image analysis model and the setting of analysis conditions of the analysis model are completed by the above-mentioned model loading / condition setting unit, initializes the analysis screen and displays an image of a normal colon, and provides the analysis results analyzed by the above-mentioned image analysis unit.

[0032] Here, the preprocessing of the colonoscopy image by the image preprocessing unit may include cropping the analysis area and adjusting the input size.

[0033] In addition, the image analysis model of the image analysis unit may be composed of a single image analysis model that detects the inspection area and the diverticulum area.

[0034] In addition, the image analysis model of the image analysis unit may be composed of an inspection area detection model that detects the inspection area and a diverticulum detection model that detects the diverticulum area.

[0035] In addition, the image analysis model of the image analysis unit may be composed of an inspection area detection model that detects the inspection area, a diverticulum detection model that detects the diverticulum area, and a lesion detection model that detects the lesion area.

[0036] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0037] In addition, the image analysis model of the image analysis unit may be composed of a cecum / diverticulum detection model that detects the cecum and diverticulum, and a lesion detection model that detects lesion sites.

[0038] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0039] In addition, when the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, the examination site may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

[0040] In addition, the control unit may transmit a command to change the indication conditions to the image analysis unit such that, when the image analysis unit detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, if it is the appendix, the cecum is indicated, and if it is a normal colon image, the diverticulum is indicated differently.

[0041] In addition, the control unit may transmit a command to the image analysis unit to change the instruction conditions so that, when the image analysis unit detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the diverticulum is not indicated in the normal colon image, or the colon diverticulum is indicated, and the examination time and retrieval time are indicated.

[0042] In addition, to achieve the above objective, a method for indicating a colonoscopy examination site according to one embodiment of the present invention is,

[0043] a) A step in which the model loading / condition setting unit loads a colonoscopy image analysis model and sets the analysis conditions of the analysis model;

[0044] b) A step in which the control unit initializes the analysis screen and displays an image of a normal colon;

[0045] c) A step in which the image preprocessing unit preprocesses the colonoscopy image received through the image receiving unit so that subsequent image analysis can be performed smoothly;

[0046] d) A step in which an image analysis unit analyzes the preprocessed colonoscopy image using an AI-based image analysis model;

[0047] e) a step in which an image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image based on the analyzed result; and

[0048] f) The control unit is characterized by including a step of providing an analysis result analyzed by the image analysis unit.

[0049] Here, the preprocessing of the colonoscopy image by the image preprocessing unit in step c) above may include cropping the analysis area and adjusting the input size.

[0050] In addition, in step d) above, the image analysis model may be composed of a single image analysis model that detects the inspection area and the diverticulum area.

[0051] In addition, in step d) above, the image analysis model may be composed of an inspection area detection model that detects an inspection area and a diverticulum detection model that detects a diverticulum area.

[0052] In addition, in step d) above, the image analysis model may be composed of an inspection area detection model that detects an inspection area, a diverticulum detection model that detects a diverticulum area, and a lesion detection model that detects a lesion area.

[0053] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0054] In addition, in step d) above, the image analysis model may be composed of a cecum / diverticulum detection model that detects the cecum and diverticulum, and a lesion detection model that detects the lesion site.

[0055] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0056] In addition, in step e) above, when the image analysis unit detects and indicates at least one of the examination site, diverticulum site, and lesion site in the colonoscopy image, the examination site may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

[0057] In addition, in step e) above, when the image analysis unit detects and indicates at least one of the examination site, diverticulum site, and lesion site in the colonoscopy image, the control unit may transmit a command to change the indication condition to the image analysis unit, such that if it is the appendix, it indicates the cecum, and if it is a normal colon image, it indicates the diverticulum differently.

[0058] In addition, in step e) above, when the image analysis unit detects and indicates at least one of the examination area, the diverticulum area, and the lesion area in the colonoscopy image, the control unit may transmit a command to change the indication conditions to the image analysis unit to prevent the diverticulum from being indicated in the normal colon image, to indicate the colon diverticulum, and to indicate the examination time and the retrieval time.

[0059] In addition, to achieve the above objective, a colonoscopy examination site indication system according to another embodiment of the present invention is,

[0060] A model loading / condition setting unit that loads a colonoscopy analysis model and a voice keyword recognition model, and sets the analysis conditions of the analysis model;

[0061] An image receiver that receives colonoscopy image frames;

[0062] An image preprocessing unit that preprocesses a colonoscopy image received through the image receiving unit to facilitate subsequent image analysis;

[0063] An image analysis unit that analyzes a colonoscopy image preprocessed by the image preprocessing unit using an AI (artificial intelligence)-based image analysis model, and detects and indicates at least one of an examination site, a diverticulum site, or a lesion site in the colonoscopy image based on the analyzed result;

[0064] A voice recognition unit that reads audio from a buffer storing audio while video analysis is being performed by the video analysis unit, analyzes it using an AI-based voice keyword recognition model, recognizes voice keywords based on the analyzed results, and transmits them to the video analysis unit; and

[0065] The invention is characterized by including a control unit that controls the status check and operation of the above-mentioned model loading / condition setting unit, image receiving unit, image preprocessing unit, image analysis unit, and voice recognition unit, and when the loading of the colonoscopy image analysis model and the setting of analysis conditions of the analysis model are completed by the above-mentioned model loading / condition setting unit, initializes the analysis screen and displays an image of a normal colon, and provides the analysis results analyzed by the above-mentioned image analysis unit, and provides the analysis results by connecting the analysis target detected by the above-mentioned image analysis model with a voice command (keyword) related to the analysis target uttered by the examiner.

[0066] Here, the preprocessing of the colonoscopy image by the image preprocessing unit may include cropping the analysis area and adjusting the input size.

[0067] In addition, the image analysis model of the image analysis unit may be composed of a single image analysis model that detects the inspection area and the diverticulum area.

[0068] In addition, the image analysis model of the image analysis unit may be composed of an inspection area detection model that detects the inspection area and a diverticulum detection model that detects the diverticulum area.

[0069] In addition, the image analysis model of the image analysis unit may be composed of an inspection area detection model that detects the inspection area, a diverticulum detection model that detects the diverticulum area, and a lesion detection model that detects the lesion area.

[0070] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0071] In addition, the image analysis model of the image analysis unit may be composed of a cecum / diverticulum detection model that detects the cecum and diverticulum, and a lesion detection model that detects lesion sites.

[0072] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0073] In addition, when the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, the examination site may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

[0074] In addition, the control unit may transmit a command to change the indication conditions to the image analysis unit such that, when the image analysis unit detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, if it is the appendix, the cecum is indicated, and if it is a normal colon image, the diverticulum is indicated differently.

[0075] In addition, the control unit may transmit a command to the image analysis unit to change the instruction conditions so that, when the image analysis unit detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the diverticulum is not indicated in the normal colon image, or the colon diverticulum is indicated, and the examination time and retrieval time are indicated.

[0076] In addition, to achieve the above objective, a method for indicating a colonoscopy examination site according to another embodiment of the present invention is,

[0077] p) A step in which the model loading / condition setting unit loads the colonoscopy analysis model and the voice keyword recognition model, and sets the analysis conditions of the analysis model;

[0078] q) A step in which the control unit initializes the analysis screen and displays an image of a normal colon;

[0079] r) A step in which the image preprocessing unit preprocesses the colonoscopy image received through the image receiving unit so that subsequent image analysis can be performed smoothly;

[0080] s) A step in which an image analysis unit analyzes a colonoscopy image preprocessed by the image preprocessing unit using an AI-based image analysis model;

[0081] t) A step in which an image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image based on the analyzed result;

[0082] u) a step in which a voice recognition unit reads audio from a buffer storing audio while video analysis is being performed by the video analysis unit, analyzes it using an AI-based voice keyword recognition model, recognizes voice keywords based on the analyzed results, and transmits them to the video analysis unit; and

[0083] v) The control unit is characterized by including a step of providing an analysis result by connecting an analysis target detected by the image analysis model of the image analysis unit with a voice command (keyword) related to an analysis target spoken by the examiner.

[0084] Here, in step r), the preprocessing of the colonoscopy image by the image preprocessing unit may include cropping the analysis area and adjusting the input size.

[0085] In addition, in step s) above, the image analysis model may be composed of a single image analysis model that detects the inspection area and the diverticulum area.

[0086] In addition, in step s) above, the image analysis model may be composed of an inspection area detection model that detects the inspection area and a diverticulum detection model that detects the diverticulum area.

[0087] In addition, in step s) above, the image analysis model may be composed of an inspection area detection model that detects the inspection area, a diverticulum detection model that detects the diverticulum area, and a lesion detection model that detects the lesion area.

[0088] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0089] In addition, in step s) above, the image analysis model may be composed of a cecum / diverticulum detection model that detects the cecum and diverticulum, and a lesion detection model that detects the lesion site.

[0090] At this time, the above lesion detection model may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0091] In addition, in step t), when the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, the examination site may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

[0092] In addition, in step t) above, when the image analysis unit detects and indicates at least one of the examination area, the diverticulum area, and the lesion area in the colonoscopy image, the control unit may transmit a command to change the indication condition to indicate the cecum if it is the appendix, and to indicate the diverticulum differently in a normal colon image.

[0093] In addition, in step t) above, when the image analysis unit detects and indicates at least one of the examination area, the diverticulum area, and the lesion area in the colonoscopy image, the control unit may transmit a command to change the indication conditions to the image analysis unit to prevent the diverticulum from being indicated in the normal colon image, to indicate the colon diverticulum, and to indicate the examination time and the retrieval time.

[0094] According to the present invention, by applying image recognition technology to the colonoscopy examination process to indicate key examination areas, the examiner is notified of situations such as diverticula protruding from the colon wall or indicated (displayed) on the examination screen differently from a normal colon, thereby allowing the examiner to proceed with the examination while paying attention to complications such as perforation, assisting in setting the probe movement path, and enabling the examiner to perform a more thorough colonoscopy examination.

[0095] In addition, by indicating the major colon examination sites, the areas examined to date can be verified; by indicating whether the cecum has been examined, it can be used as an indicator of colonoscopy quality; and by recording the presence of diverticula, it has the advantage of allowing the examiner to be aware of the risk of perforation in advance and proceed with the next colonoscopy.

[0096] FIG. 1 is a schematic diagram showing the configuration of a colonoscopy examination site indication system according to one embodiment of the present invention.

[0097] FIGS. 2a to 2d are drawings showing examples of configurations of image analysis models.

[0098] FIG. 3 is a flowchart illustrating the execution process of a method for indicating a colonoscopy examination site according to one embodiment of the present invention.

[0099] Figure 4 is a flowchart showing the execution process of a first modified example of the colonoscopy examination site indication method of Figure 3.

[0100] Figure 5 is a flowchart showing the execution process of a second modified example of the colonoscopy examination site indication method of Figure 3.

[0101] Figure 6 is a flowchart showing the execution process of a third modified example of the colonoscopy examination site indication method of Figure 3.

[0102] FIG. 7 is a schematic diagram showing the configuration of a colonoscopy examination site indication system according to another embodiment of the present invention.

[0103] FIG. 8 is a flowchart illustrating the execution process of a method for indicating a colonoscopy examination site according to another embodiment of the present invention.

[0104] FIG. 9 is a flowchart showing the execution process of a modified example of a method for indicating a colonoscopy examination site according to another embodiment of the present invention.

[0105] Figure 10 is a diagram showing the major parts of the large intestine.

[0106] Figure 11 is a diagram showing an example of a ledger marking.

[0107] Figure 12 is a diagram showing the inspection area of ​​the large intestine (in the case of a normal large intestine).

[0108] Figure 13 is a diagram showing the inspection site of the large intestine (in the case of a left large intestine diverticulum).

[0109] Figure 14 is a diagram showing the inspection site of the large intestine (in the case of left and right large intestine diverticula).

[0110] Figure 15 is a diagram showing another example of an examination site indication of the large intestine (in the case of a left colonic diverticulum).

[0111] Figure 16 is a diagram showing the inspection site of the large intestine (in the case of right and left colonic diverticula).

[0112] Figure 17 is a diagram showing the inspection area of ​​the large intestine (indicating whether diverticula are found by distinguishing between the left and right sides).

[0113] Figure 18 is a diagram showing an example of providing analysis results (an example of a normal colon examination).

[0114] Figure 19 is a diagram showing an example of providing analysis results (an example of finding a diverticulum in the cecum).

[0115] Figure 20 is a diagram showing an example of providing analysis results (an example of finding a diverticulum in the sigmoid colon).

[0116] Figure 21 is a diagram showing an example of providing analysis results (example of a result report).

[0117] Embodiments of the present invention will be described in detail below with reference to the attached drawings.

[0118] FIG. 1 is a schematic diagram showing the configuration of a colonoscopy examination site indication system according to one embodiment of the present invention.

[0119] Referring to FIG. 1, a colonoscopy examination site indication system (100) according to one embodiment of the present invention may be configured to include a model loading / condition setting unit (110), an image receiving unit (120), an image preprocessing unit (130), an image analysis unit (140), and a control unit (150).

[0120] The model loading / condition setting unit (110) loads a colonoscopy image analysis model and sets the analysis conditions of the analysis model. Here, the analysis conditions of the analysis model can be set, for example, to a predicted probability value of 0.85 or higher.

[0121] The image receiving unit (120) receives a colonoscopy image frame.

[0122] The image preprocessing unit (130) preprocesses the colonoscopy image received through the image receiving unit (120) so that subsequent image analysis can be performed smoothly. Here, the preprocessing of the colonoscopy image by the image preprocessing unit (130) as described above may include cutting the analysis area and adjusting the input size.

[0123] The image analysis unit (140) analyzes the colonoscopy image preprocessed by the image preprocessing unit (130) using an AI (artificial intelligence)-based image analysis model, and based on the analyzed results, detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image. Here, the image analysis model of the image analysis unit (140) may be composed of a single image analysis model that detects the examination area and diverticulum area, as shown in FIG. 2a. In FIG. 2a, (a) shows an example of internal indication from the entry into the colon, and (b) shows an example of cecum-centered indication.

[0124] In addition, the image analysis model of the image analysis unit (140) may be composed of an inspection area detection model (Model A) that detects the inspection area and a diverticulum detection model (Model B) that detects the diverticulum area, as shown in FIG. 2b. In FIG. 2b, (a) shows an example of internal indication from entry into the large intestine, and (b) shows an example of cecum center indication.

[0125] In addition, the image analysis model of the image analysis unit (140) may be composed of an inspection area detection model (Model A) that detects the inspection area, a diverticulum detection model (Model B) that detects the diverticulum area, and a lesion detection model (Model C) that detects the lesion area, as shown in FIG. 2C. At this time, the lesion detection model (Model C) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0126] In addition, the image analysis model of the image analysis unit (140) may be composed of a cecum / diverticulum detection model (Model A) that detects the cecum and diverticulum and a lesion detection model (Model B) that detects the lesion area, as shown in FIG. 2D. At this time, the lesion detection model (Model B) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0127] In addition, when the image analysis unit (140) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the examination area may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum, as shown in FIGS. 2a to 2c.

[0128] The control unit (150) controls the status check and operation of the model loading / condition setting unit (110), image receiving unit (120), image preprocessing unit (130), and image analysis unit (140). When the loading of the colonoscopy image analysis model and the setting of analysis conditions for the analysis model are completed by the model loading / condition setting unit (110), the control unit (150) initializes the analysis screen, displays an image of a normal colon, and provides the analysis results analyzed by the image analysis unit (140). Here, the control unit (150) can transmit a command to change the instruction conditions to the image analysis unit (140) such that when the image analysis unit (140) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, if it is the appendix, it indicates the cecum, and if it is a normal colon image, it indicates the diverticulum differently.

[0129] Additionally, the control unit (150) can transmit a command to the image analysis unit (140) to change the instruction conditions so that, when the image analysis unit (140) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the diverticulum is not indicated in the normal colon image, or the colon diverticulum is indicated (e.g., alarm or area indication), and the examination time and recovery time are indicated.

[0130] In FIG. 1, reference number 160 represents a database (DB), and such a database (DB) (160) stores and manages various software programs for system operation, data or information required when the model loading / condition setting unit (110), image receiving unit (120), image preprocessing unit (130), and image analysis unit (140) perform functions related to model loading and condition setting, image preprocessing, and image analysis, or process tasks, as well as colonoscopy image analysis result data based on an image analysis model.

[0131] Here, the model loading / condition setting unit (110), image receiving unit (120), image preprocessing unit (130), image analysis unit (140), control unit (150), and database (DB) (160) as described above may also be integrated as a whole to form a single computer system.

[0132] Then, below, we will describe a method for indicating a colonoscopy examination site based on a colonoscopy examination site indicating system according to one embodiment of the present invention having the configuration as described above.

[0133] FIG. 3 is a flowchart illustrating the execution process of a method for indicating a colonoscopy examination site according to one embodiment of the present invention.

[0134] Referring to FIG. 3, a method for indicating a colonoscopy examination site according to one embodiment of the present invention first loads a colonoscopy image analysis model by a model loading / condition setting unit (110) and sets an analysis condition of the analysis model (e.g., a predicted probability value of 0.85 or higher) (step S301).

[0135] Then, the control unit (150) initializes the analysis screen and displays a picture of a normal colon (step S302).

[0136] As described above, after the model loading and analysis condition settings are completed, the analysis screen is initialized and a normal colon image is displayed, the control unit (150) determines whether to perform colonoscopy image analysis (step S303). If colonoscopy image analysis is required during this determination, the image preprocessing unit (130) reads the colonoscopy image (image frame) received through the image receiving unit (120) and preprocesses it so that subsequent image analysis can be performed smoothly (step S304). Here, the preprocessing of the colonoscopy image by the image preprocessing unit (130) may include cutting the analysis area and adjusting the input size.

[0137] When the preprocessing of the colonoscopy image is completed in this way, the image analysis unit (140) analyzes the preprocessed colonoscopy image using an AI-based image analysis model (step S305). Here, the image analysis model may be composed of a single image analysis model that detects the examination area and the diverticulum area as shown in FIG. 2a, as described above. In addition, the image analysis model may be composed of an examination area detection model (Model A) that detects the examination area and a diverticulum detection model (Model B) that detects the diverticulum area, as shown in FIG. 2b. In addition, the image analysis model may be composed of an examination area detection model (Model A) that detects the examination area, a diverticulum detection model (Model B) that detects the diverticulum area, and a lesion detection model (Model C) that detects the lesion area, as shown in FIG. 2c. At this time, the lesion detection model (Model C) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant. In addition, as illustrated in FIG. 2d, the image analysis model may be composed of a cecum / diverticulum detection model (Model A) that detects the cecum and diverticulum, and a lesion detection model (Model B) that detects lesion sites. At this time, the lesion detection model (Model B) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0138] Additionally, the image analysis unit (140) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image based on the analyzed result (steps S306 to S308). Here, when the image analysis unit (140) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the examination area may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

[0139] Here, we will explain the steps S306 to S308 above in a little more detail.

[0140] When the colonoscopy image analysis is completed by the image analysis unit (140) in step S305, the control unit (150) determines whether a diverticulum is detected (step S306), and if a diverticulum is detected, changes the normal colon image to a diverticulum colon image and displays it (see FIGS. 13–16) (step S307), and indicates the examination area on the colon image (step S308).

[0141] Afterward, if there is no further request for colonoscopy image analysis in the determination of step S303, the control unit (150) provides the analysis result analyzed immediately prior to by the image analysis unit (140) (step S309).

[0142] Meanwhile, Fig. 4 is a flowchart showing the execution process of a first modified example of the colonoscopy examination site indication method of Fig. 3.

[0143] Referring to FIG. 4, the process described above in FIG. 3 differs from FIG. 3 only in that additional processes are added: storing the inspection start time (step S404), storing the appendix instruction start time (step S410), displaying the inspection time and retrieval time (step S411), determining the instruction condition change (step S412), changing the instruction condition (step S413), and storing the inspection end time (step S414). The rest of the process is identical to FIG. 3. Therefore, the description of the parts identical to FIG. 3 (steps S401–S403, S405–S409, S415) will be replaced by the description in FIG. 3 (i.e., the description of steps S301–S303, S304–S308, S309), and only the parts different from FIG. 3 will be described.

[0144] If colonoscopy image analysis is required in the determination of step S403 of Fig. 4, the control unit (150) stores the examination start time in the database (160) (step S404).

[0145] Afterwards, as described in FIG. 3, the image preprocessing unit (130) reads the colonoscopy image (image frame) received through the image receiving unit (120) and preprocesses it so that subsequent image analysis can be performed smoothly (step S405).

[0146] Additionally, the control unit (150) indicates the inspection area on the colon diagram (step S409) and stores the start time of the cecum indication (step S410). Then, the control unit (150) displays the inspection time and the retrieval time (step S411). Here, the inspection time refers to the time from the start time of the inspection to the current time (end of inspection), and the retrieval time refers to the time from the start time of the cecum indication to the current time (end of inspection).

[0147] Additionally, the control unit (150) determines whether to change the instruction condition (step S412), and if a change in the instruction condition is required, changes the instruction condition (step S413). Here, the control unit (150) can transmit an instruction condition change command to the image analysis unit (140) such that if it is the appendix, it indicates the cecum, and if it is a normal colon image, it indicates the diverticulum differently.

[0148] Additionally, the control unit (150) can transmit a command to change the instruction conditions to the image analysis unit (140) to prevent the diverticulum from being indicated in the normal colon image, to indicate the colon diverticulum (e.g., alarm or area indication), and to indicate the inspection time and recovery time.

[0149] Meanwhile, if there is no further request for colonoscopy image analysis in the determination of step S403, the control unit (150) stores the time of end of the examination (step S414) and provides the analysis results up to the present (end of examination) (step S415).

[0150] Figure 5 is a flowchart showing the execution process of a second modified example of the colonoscopy examination site indication method of Figure 3.

[0151] Referring to FIG. 5, the process described above in FIG. 3 differs from FIG. 3 only in that a step (S509) for determining whether a lesion is detected and a step (S510) for indicating lesion information on a colon image are added, while the rest of the process is identical to FIG. 3. Therefore, the description of the parts identical to FIG. 3 here will be replaced with the description in FIG. 3, and only the parts different from FIG. 3 will be described.

[0152] In FIG. 5, the control unit (150) indicates the inspection area on the colon drawing (step S508), determines whether a lesion is detected (step S509), and if a lesion is detected, indicates the lesion information on the colon drawing (step S510).

[0153] Figure 6 is a flowchart showing the execution process of a third modified example of the colonoscopy examination site indication method of Figure 3.

[0154] Referring to FIG. 6, the process described above in FIG. 4 is different from FIG. 4 only in that a step (S611) for determining whether a lesion is detected and a step (S612) for indicating lesion information in a colon drawing are added, while the rest of the process is the same as FIG. 4.

[0155] That is, if colonoscopy image analysis is required in the determination of step S603 of FIG. 6, the control unit (150) stores the examination start time in the database (160) (step S604).

[0156] Afterwards, as described in FIG. 3, the image preprocessing unit (130) reads the colonoscopy image (image frame) received through the image receiving unit (120) and preprocesses it so that subsequent image analysis can be performed smoothly (step S605).

[0157] Additionally, the control unit (150) indicates the inspection area on the colon diagram (step S609) and stores the start time of the appendix indication (step S610). Then, the control unit (150) determines whether a lesion is detected (step S611), and if a lesion is detected, indicates the lesion information on the colon diagram (step S612).

[0158] Then, the control unit (150) displays the inspection time and the retrieval time (step S613). Here, the inspection time refers to the time from the start time of the inspection to the current time (end of inspection), and the retrieval time refers to the time from the start time of the appendix instruction to the current time (end of inspection).

[0159] Additionally, the control unit (150) determines whether to change the instruction condition (step S614), and if a change in the instruction condition is required, changes the instruction condition (step S615). Here, the control unit (150) can transmit an instruction condition change command to the image analysis unit (140) to indicate the cecum if it is the appendix, and to indicate the diverticulum differently in the normal colon illustration.

[0160] Additionally, the control unit (150) can transmit a command to change the instruction conditions to the image analysis unit (140) to prevent the diverticulum from being indicated in the normal colon image, to indicate the colon diverticulum (e.g., alarm or area indication), and to indicate the inspection time and recovery time.

[0161] Meanwhile, if there is no further request for colonoscopy image analysis in the determination of step S603, the control unit (150) stores the time of end of the examination (step S616) and provides the analysis results up to the present (end of examination) (step S617).

[0162] FIG. 7 is a schematic diagram showing the configuration of a colonoscopy examination site indication system according to another embodiment of the present invention.

[0163] Referring to FIG. 7, a colonoscopy examination site indication system (700) according to another embodiment of the present invention basically has the same components as the colonoscopy examination site indication system (100) according to one embodiment described above with reference to FIG. 1. However, the colonoscopy examination site indication system (700) according to this other embodiment differs in that it further includes a voice recognition unit (750).

[0164] As illustrated in FIG. 7, a colonoscopy examination site indication system (700) according to another embodiment of the present invention may be configured to include a model loading / condition setting unit (710), an image receiving unit (720), an image preprocessing unit (730), an image analysis unit (740), a voice recognition unit (750), and a control unit (760).

[0165] The model loading / condition setting unit (710) loads a colonoscopy image analysis model and sets the analysis conditions of the analysis model. Here, the analysis conditions of the analysis model can be set, for example, to a predicted probability value of 0.85 or higher.

[0166] The image receiving unit (720) receives a colonoscopy image frame.

[0167] The image preprocessing unit (730) preprocesses the colonoscopy image received through the image receiving unit (720) so that subsequent image analysis can be performed smoothly. Here, the preprocessing of the colonoscopy image by the image preprocessing unit (730) as described above may include cutting the analysis area and adjusting the input size.

[0168] The image analysis unit (740) analyzes the colonoscopy image preprocessed by the image preprocessing unit (730) using an AI (artificial intelligence)-based image analysis model, and based on the analyzed results, detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image. Here, the image analysis model of the image analysis unit (740) may be composed of a single image analysis model that detects the examination area and diverticulum area, as shown in FIG. 2a. In FIG. 2a, (a) shows an example of internal indication from the entry into the colon, and (b) shows an example of cecum-centered indication.

[0169] In addition, the image analysis model of the image analysis unit (740) may be composed of an inspection area detection model (Model A) that detects the inspection area and a diverticulum detection model (Model B) that detects the diverticulum area, as shown in FIG. 2b. In FIG. 2b, (a) shows an example of internal indication from the entry into the large intestine, and (b) shows an example of cecum-centered indication.

[0170] In addition, the image analysis model of the image analysis unit (740) may be composed of an inspection area detection model (Model A) that detects the inspection area, a diverticulum detection model (Model B) that detects the diverticulum area, and a lesion detection model (Model C) that detects the lesion area, as illustrated in FIG. 2C. At this time, the lesion detection model (Model C) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0171] In addition, the image analysis model of the image analysis unit (740) may be composed of a cecum / diverticulum detection model (Model A) that detects the cecum and diverticulum and a lesion detection model (Model B) that detects the lesion area, as shown in FIG. 2D. At this time, the lesion detection model (Model B) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0172] In addition, when the image analysis unit (740) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the examination area may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

[0173] The voice recognition unit (750) reads audio from a buffer (located in the internal memory of the voice recognition unit (750)) that stores audio while video analysis is being performed by the video analysis unit (740), analyzes it using an AI-based voice keyword recognition model, recognizes voice keywords based on the analyzed results, and transmits them to the video analysis unit (740).

[0174] The control unit (760) controls the status check and operation of the model loading / condition setting unit (710), image receiving unit (720), image preprocessing unit (730), image analysis unit (740), and voice recognition unit (750). When the loading of the colonoscopy image analysis model and the setting of the analysis conditions of the analysis model are completed by the model loading / condition setting unit (710), the analysis screen is initialized and an image of a normal colon is displayed. The analysis results analyzed by the image analysis unit (740) are provided by connecting the analysis target detected by the image analysis model with a voice command (keyword) related to the analysis target spoken by the examiner.

[0175] Here, the control unit (760) described above can transmit a command to change the instruction condition to the image analysis unit (740) such that when the image analysis unit (740) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, if it is the appendix, it indicates the cecum, and if it is a normal colon image, it indicates the diverticulum differently.

[0176] Additionally, the control unit (760) can transmit a command to the image analysis unit (740) to change the instruction conditions so that, when the image analysis unit (740) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the diverticulum is not indicated in the normal colon image, or the colon diverticulum is indicated (e.g., alarm or area indication), and the examination time and recovery time are indicated.

[0177] In FIG. 7, reference number 770 represents a database (DB), and such a database (DB) (770) stores and manages various software programs for system operation, data or information required when the model loading / condition setting unit (710), image receiving unit (720), image preprocessing unit (730), image analysis unit (740), and voice recognition unit (750) perform functions related to model loading and condition setting, image preprocessing, image analysis, and voice recognition, or process tasks, as well as colonoscopy image analysis result data based on an image analysis model.

[0178] Here, the model loading / condition setting unit (710), image receiving unit (720), image preprocessing unit (730), image analysis unit (740), voice recognition unit (750), control unit (760), and database (DB) (770) as described above may also be integrated as a whole to form a single computer system.

[0179] Then, below, we will describe a method for indicating a colonoscopy examination site based on a colonoscopy examination site indicating system according to another embodiment of the present invention having the configuration as described above.

[0180] FIG. 8 is a flowchart illustrating the execution process of a method for indicating a colonoscopy examination site according to another embodiment of the present invention.

[0181] Referring to FIG. 8, a method for indicating a colonoscopy examination site according to another embodiment of the present invention first loads a colonoscopy image analysis model and a voice recognition model through a model loading / condition setting unit (710), and sets an analysis condition of the analysis model (e.g., a predicted probability value of 0.85 or higher) (step S801).

[0182] Then, the control unit (760) initializes the analysis screen and displays a picture of a normal colon (step S802).

[0183] As described above, after the loading of the colonoscopy image analysis model, the loading of the voice recognition model, and the setting of the analysis conditions of the models are completed, and after the analysis screen is initialized and a normal colon image is displayed, the control unit (760) determines whether to perform colonoscopy image analysis (step S803). If colonoscopy image analysis is required in this determination, the image preprocessing unit (730) reads the colonoscopy image (image frame) received through the image receiving unit (720) and preprocesses it so that subsequent image analysis can be performed smoothly (step S804). Here, the preprocessing of the colonoscopy image by the image preprocessing unit (730) may include cutting the analysis area and adjusting the input size.

[0184] When the preprocessing of the colonoscopy image is completed in this way, the image analysis unit (740) analyzes the preprocessed colonoscopy image using an AI-based image analysis model (step S805). Here, the image analysis model may be composed of a single image analysis model that detects the examination area and the diverticulum area as shown in FIG. 2a, as described above. Additionally, the image analysis model may be composed of an examination area detection model (Model A) that detects the examination area and a diverticulum detection model (Model B) that detects the diverticulum area, as shown in FIG. 2b. Additionally, the image analysis model may be composed of an examination area detection model (Model A) that detects the examination area, a diverticulum detection model (Model B) that detects the diverticulum area, and a lesion detection model (Model C) that detects the lesion area, as shown in FIG. 2c. At this time, the lesion detection model (Model C) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant. In addition, the image analysis model may be composed of a cecum / diverticulum detection model (Model A) that detects the cecum and diverticulum, and a lesion detection model (Model B) that detects lesion sites, as shown in FIG. 2D. At this time, the lesion detection model (Model B) may be equipped with a lesion attribute classification function that distinguishes whether the lesion is benign or malignant.

[0185] Additionally, the image analysis unit (740) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image based on the analyzed result (steps S806 to S808). Here, when the image analysis unit (740) detects and indicates at least one of the examination area, diverticulum area, and lesion area in the colonoscopy image, the examination area may include the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

[0186] Here, we will explain the above steps S806 to S808 in a little more detail.

[0187] When the colonoscopy image analysis is completed by the image analysis unit (740) in step S805, the control unit (760) determines whether a diverticulum is detected (step S806), and if a diverticulum is detected, changes the normal colon image to a diverticulum colon image and displays it (see FIGS. 13–16) (step S807), and indicates the examination area on the colon image (step S808).

[0188] Meanwhile, if colonoscopy image analysis is required during the determination of step S803, the voice recognition unit (750) reads audio from a buffer storing audio while image analysis is being performed by the image analysis unit (740), analyzes it using an AI-based voice keyword recognition model (step S809), recognizes voice keywords based on the analyzed results (step S810), and transmits them to the image analysis unit (740).

[0189] That is, the voice recognition unit (750) recognizes a voice keyword and determines whether the voice keyword is a diverticulum keyword (step S811), and if it is a diverticulum keyword, transmits it to step S807 so that it is reflected in changing to a diverticulum colon image.

[0190] And, if the keyword is not diverticulum in the above determination, the voice recognition unit (750) determines whether it is an examination area keyword (step S812), and if it is an examination area keyword, transmits it to step S808 so that it is reflected in indicating the examination area on the colon drawing.

[0191] Subsequently, if there is no further request for colonoscopy image analysis in the determination of step S803, the control unit (760) connects the analysis target detected by the image analysis model of the image analysis unit (740) with the voice command (keyword) related to the analysis target uttered by the examiner, and provides the analysis results up to the present (end of examination) by the image analysis unit (740) (step S813).

[0192] FIG. 9 is a flowchart showing the execution process of a modified example of a method for indicating a colonoscopy examination site according to another embodiment of the present invention.

[0193] Referring to FIG. 9, the process described above in FIG. 8 is different from FIG. 8 only in that additional processes are added: storing the start time of the inspection (step S904), storing the start time of the appendix indication (step S910), determining whether a lesion is detected (step S911), indicating lesion information on the colon diagram (step S912), displaying the inspection time and retrieval time (step S913), determining whether the indication condition is changed (step S914), changing the indication condition (step S915), and storing the end time of the inspection (step S920). The rest of the process is identical to FIG. 8. Therefore, the description of the parts identical to FIG. 8 will be replaced by the description in FIG. 8, and only the parts different from FIG. 8 will be described.

[0194] If colonoscopy image analysis is required in the determination of step S903 of Fig. 9, the control unit (760) stores the examination start time in the database (770) (step S904).

[0195] Afterwards, as described in FIG. 8, the image preprocessing unit (730) reads the colonoscopy image (image frame) received through the image receiving unit (720) and preprocesses it so that subsequent image analysis can be performed smoothly (step S905).

[0196] Additionally, the control unit (760) indicates the inspection area on the colon diagram (step S909) and stores the start time of the appendix indication (step S910). Then, the control unit (760) determines whether a lesion is detected (step S911), and if a lesion is detected, indicates the lesion information on the colon diagram (step S912).

[0197] Then, the control unit (760) displays the inspection time and the retrieval time (step S913). Here, the inspection time refers to the time from the start time of the inspection to the current time (end of inspection), and the retrieval time refers to the time from the start time of the appendix instruction to the current time (end of inspection).

[0198] Additionally, the control unit (760) determines whether to change the instruction condition (step S914), and if a change in the instruction condition is required, changes the instruction condition (step S915). Here, the control unit (760) can transmit an instruction condition change command to the image analysis unit (740) to indicate the cecum if it is the appendix, and to indicate the diverticulum differently in the normal colon illustration.

[0199] Additionally, the control unit (760) can transmit a command to change the instruction conditions to the image analysis unit (740) to prevent the diverticulum from being indicated in the normal colon image, to indicate the colon diverticulum (e.g., alarm or area indication), and to indicate the inspection time and recovery time.

[0200] Meanwhile, if there is no further request for colonoscopy image analysis in the determination of step S903, the control unit (760) stores the time of end of the examination (step S920) and provides the analysis results up to the present (end of examination) (step S921).

[0201] Below, we will provide further explanation regarding the colonoscopy examination site indication system and method according to the present invention as described above.

[0202] Figure 10 is a diagram showing the major parts of the large intestine.

[0203] Referring to FIG. 10, (a) shows the main part of the large intestine, (b) shows the examination sequence of the probe insertion process (direction of colonoscopy photography), and (c) shows the examination sequence of the probe retrieval process.

[0204] Referring to (a) and (b), the colonoscope is inserted through the rectum and enters the large intestine. Once inside the large intestine, the colonoscope photographs the inside of the large intestine in the order of the sigmoid colon → descending colon → transverse colon → ascending colon → cecum → appendix.

[0205] Referring to (a) and (c), this is an examination of the probe retrieval process. Once inserted to the end of the large intestine, the colonoscope is retracted in reverse to image the inside of the large intestine in the order of appendix → cecum → ascending colon → transverse colon → descending colon → sigmoid colon → rectum. In (a) of Fig. 10, the ileocecal valve indicates the junction between the cecum and the small intestine.

[0206] In the above series of colonoscopy probe movement and imaging, when the probe is inserted to position 8 during the probe insertion process of (b), it is considered that the cecum has been confirmed. The examination is then carried out by retrieving the probe after confirming the cecum. Professional examiners also perform observation while inserting the probe.

[0207] Figure 11 is a diagram showing an example of a ledger marking.

[0208] Referring to Fig. 11, (a) shows a normal colon, (b) shows a left colonic diverticulum, and (c) shows a right colonic diverticulum. The left colonic diverticulum in (b) mainly occurs in the sigmoid colon and descending colon, while the right colonic diverticulum in (c) mainly occurs in the cecum.

[0209] Here, diverticula as described above are caused by increased pressure within the large intestine. The diverticula in the left colon (pseudodiverticula) shown in (b) appear as multiple diverticula in the left colon, with a portion of the intestinal wall (mucosal and submucosal tissues) protruding; these are acquired and common in Westerners, but recently, there has been an increasing trend among Asians as well. The diverticula in the right colon (true diverticula) shown in (c) appear as a single diverticulum in the right colon, with the entire intestinal wall, including the muscle layer, protruding; these are congenital and common in Asians.

[0210] Figure 12 is a diagram showing the inspection area of ​​the large intestine (in the case of a normal large intestine).

[0211] Referring to Fig. 12, this shows a case where the examination site is indicated during the colonoscopy probe insertion process. Generally, the inside of the large intestine is examined (inspected) in the order of examination start → rectal examination → sigmoid colon examination → descending colon examination → transverse colon examination → ascending colon examination → cecum examination.

[0212] Figure 13 is a diagram showing the inspection site of the large intestine (in the case of a left large intestine diverticulum).

[0213] Referring to Fig. 13, this shows a case where the examination site is indicated during the process of inserting a colonoscopy probe. Similar to the case of a normal colon in Fig. 12, the inside of the colon is examined in the order of examination start → rectal observation → sigmoid colon observation → descending colon observation → transverse colon observation → ascending colon observation → cecum observation. This example shows a case where a diverticulum is found in the sigmoid colon.

[0214] Figure 14 is a diagram showing the inspection site of the large intestine (in the case of left and right large intestine diverticula).

[0215] Referring to Fig. 14, this shows a case where the examination site is indicated during the process of inserting a colonoscopy probe. Likewise, the inside of the large intestine is observed (examined) in the order of examination start → rectal observation → sigmoid colon observation → descending colon observation → transverse colon observation → ascending colon observation → cecum observation. This example shows a case where diverticula were found in the sigmoid colon, descending colon, and cecum.

[0216] Figure 15 is a diagram showing another example of an examination site indication of the large intestine (in the case of a left colonic diverticulum).

[0217] Referring to Fig. 15, this shows a case where the examination site is indicated during the process of inserting the colonoscopy probe. Likewise, the inside of the colon is observed (examined) in the order of examination start → rectal observation → sigmoid colon observation → descending colon observation → transverse colon observation → ascending colon observation → cecum observation. In this example, only the insertion into the cecum is confirmed, and a diverticulum is found in the sigmoid colon.

[0218] Figure 16 is a diagram showing the inspection site of the large intestine (in the case of right and left colonic diverticula).

[0219] Referring to Fig. 16, this shows a case where the examination area is indicated during the colonoscopy probe retrieval process, and the inside of the colon is observed (examined) in the order of examination start → cecum observation → ascending colon observation → transverse colon observation → descending colon observation → sigmoid colon observation → rectum observation (i.e., while retrieving the probe in reverse).

[0220] Figure 17 is a diagram showing the inspection area of ​​the large intestine (indicating whether diverticula are found by distinguishing between the left and right sides).

[0221] Referring to FIG. 17, this shows a case where the presence of diverticula is briefly indicated by distinguishing only between the left and right colons. As illustrated, in the case of a right colon diverticulum, a left colon diverticulum, or both right and left colon diverticula, a specific area of ​​the corresponding left or right, or left / right colon is indicated (marked) so that the examiner can visually confirm it. In FIG. 17, (a) indicates the indication of the main examination area, and (b) indicates the confirmation of whether the cecum has been inserted.

[0222] Figure 18 is a diagram showing an example of providing analysis results (an example of a normal colon examination).

[0223] Referring to FIG. 18, this illustrates the provision of analysis results for a normal colonoscopy, wherein (a) shows the colonoscopy start screen, (b) shows an example screen of a cecum examination during the probe insertion process, and (c) shows an example screen of a cecum examination during the probe retrieval process. In FIG. 18, reference number 810 represents the colonoscopy image analysis software screen, 820 represents the colonoscopy image analysis area, 830 represents the ileocecal valve, 840 represents the appendix, and 850 represents the cecum. Additionally, T represents the examination time and W represents the retrieval time.

[0224] Figure 19 is a diagram showing an example of providing analysis results (an example of finding a diverticulum in the cecum).

[0225] Referring to FIG. 19, this illustrates the analysis results of the discovery of diverticula in the cecum. Similarly, (a) represents the colonoscopy start screen, (b) represents an example screen of the cecum examination during the probe insertion process, and (c) represents an example screen of the cecum examination during the probe retrieval process. In FIG. 19, reference number 810 represents the colonoscopy image analysis software screen, 820 represents the colonoscopy image analysis area, 830 represents the ileocecal valve, 840 represents the appendix, 850 represents the cecum, and 860 represents the diverticulum. Additionally, T represents the examination time and W represents the retrieval time.

[0226] Figure 20 is a diagram showing an example of providing analysis results (an example of finding a diverticulum in the sigmoid colon).

[0227] Referring to FIG. 20, this illustrates the analysis results of the discovery of diverticula in the sigmoid colon. Similarly, (a) shows the colonoscopy start screen, (b) shows an example of a sigmoid colon examination during the probe insertion process, and (c) shows an example of a sigmoid colon examination during the probe retrieval process. In FIG. 20, reference number 810 represents the colonoscopy image analysis software screen, 860 represents the diverticulum, and 870 represents the sigmoid colon. Additionally, T represents the examination time and W represents the retrieval time.

[0228] Figure 21 is a diagram showing an example of providing analysis results (example of a result report).

[0229] Referring to FIG. 21, the colonoscopy results are shown as follows: (a) indicates the examination site from probe insertion, showing the case where a cecal examination is performed and there are no colonic diverticula; (b) indicates whether a cecal examination is performed, showing the case where a cecal examination is performed and there are no colonic diverticula; (c) indicates whether a cecal examination is performed and a diverticula are present, showing the case where a cecal examination is performed and there are colonic diverticula; and (d) indicates whether a cecal examination is performed and a brief indication of left colonic diverticula, showing the case where a cecal examination is performed and there are colonic diverticula in the left colon.

[0230] As described above, the colonoscopy examination site indication system and method according to the present invention applies image recognition technology to the colonoscopy examination process to indicate key examination sites. By notifying the examiner of situations such as diverticula protruding from the colon wall or indicating (displaying) them on the examination screen differently from a normal colon, the examiner can proceed with the examination while paying attention to complications such as perforation, assist in setting the probe movement path, and enable the examiner to perform a more thorough colonoscopy examination.

[0231] In addition, by indicating the major colon examination sites, the areas examined to date can be verified; by indicating whether the cecum has been examined, it can be used as an indicator of colonoscopy quality; and by recording the presence of diverticula, it has the advantage of allowing the examiner to be aware of the risk of perforation in advance and proceed with the next colonoscopy.

[0232] Although the present invention has been described in detail through preferred embodiments, the invention is not limited thereto, and it is obvious to those skilled in the art that various modifications and applications can be made within the scope of the technical concept of the invention. Accordingly, the true scope of protection of the present invention should be interpreted by the following claims, and all technical concepts within an equivalent scope should be interpreted as being included within the scope of rights of the present invention.

Claims

1. Model loading / condition setting unit for loading a colonoscopy image analysis model and setting analysis conditions for the analysis model; An image receiver that receives colonoscopy image frames; An image preprocessing unit that preprocesses a colonoscopy image received through the image receiving unit to facilitate subsequent image analysis; An image analysis unit that analyzes a colonoscopy image preprocessed by the image preprocessing unit using an AI (artificial intelligence)-based image analysis model, and detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image based on the analyzed result; and A colonoscopy examination site indication system comprising a control unit that controls the status check and operation of the above-mentioned model loading / condition setting unit, image receiving unit, image preprocessing unit, and image analysis unit, and when the loading of the colonoscopy image analysis model and the setting of analysis conditions of the analysis model are completed by the above-mentioned model loading / condition setting unit, initializes the analysis screen and displays an image of a normal colon, and provides an analysis result analyzed by the above-mentioned image analysis unit.

2. In Paragraph 1, Preprocessing of colonoscopy images by the above-mentioned image preprocessing unit is a colonoscopy examination site indication system including analysis area cutting and input size adjustment.

3. In Paragraph 1, A colonoscopy examination site indication system characterized in that the image analysis model of the above-mentioned image analysis unit is composed of a single image analysis model that detects the examination site and the diverticulum site.

4. In Paragraph 1, A colonoscopy examination site indication system characterized in that the image analysis model of the above-mentioned image analysis unit is composed of an examination site detection model that detects the examination site and a diverticulum detection model that detects the diverticulum site.

5. In Paragraph 1, A colonoscopy examination site indication system characterized in that the image analysis model of the image analysis unit is composed of an examination site detection model that detects the examination site, a diverticulum detection model that detects the diverticulum site, and a lesion detection model that detects the lesion site.

6. In Paragraph 5, A colonoscopy examination site indication system characterized by the above-mentioned lesion detection model having a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

7. In Paragraph 1, A colonoscopy examination site indication system characterized by the image analysis model of the above-mentioned image analysis unit being composed of a cecum / diverticulum detection model that detects the cecum and diverticulum and a lesion detection model that detects the lesion site.

8. In Paragraph 7, A colonoscopy examination site indication system characterized by the above-mentioned lesion detection model having a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

9. In Paragraph 1, A colonoscopy examination site indicating system in which the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, wherein the examination site includes the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

10. In Paragraph 1, A colonoscopy examination site indication system characterized by the above-described control unit transmitting an indication condition change command to the above-described image analysis unit, wherein the above-described image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, such that if it is the appendix, the cecum is indicated, and if it is a normal colon image, the diverticulum is indicated differently.

11. In Paragraph 1, A colonoscopy examination site indication system characterized by the above-described control unit transmitting an instruction condition change command to the above-described image analysis unit, wherein the above-described image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, such that the diverticulum is not indicated in the normal colon illustration, the colon diverticulum is indicated, and the examination time and retrieval time are indicated. 12.a) A step in which the model loading / condition setting unit loads a colonoscopy image analysis model and sets the analysis conditions of the analysis model; b) A step in which the control unit initializes the analysis screen and displays an image of a normal colon; c) A step in which the image preprocessing unit preprocesses the colonoscopy image received through the image receiving unit so that subsequent image analysis can be performed smoothly; d) A step in which an image analysis unit analyzes the preprocessed colonoscopy image using an AI-based image analysis model; e) a step in which an image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image based on the analyzed result; and f) A method for indicating a colonoscopy examination site comprising the step of a control unit providing an analysis result analyzed by the image analysis unit.

13. In Paragraph 12, The method for indicating a colonoscopy examination site, wherein the preprocessing of the colonoscopy image by the image preprocessing unit in step c) above includes cutting the analysis area and adjusting the input size.

14. In Paragraph 12, A method for indicating a colonoscopy examination site, characterized in that, in step d) above, the image analysis model is composed of a single image analysis model that detects the examination site and the diverticulum site.

15. In Paragraph 12, A method for indicating a colonoscopy examination site, characterized in that, in step d) above, the image analysis model is composed of an examination site detection model that detects an examination site and a diverticulum detection model that detects a diverticulum site.

16. In Paragraph 12, A method for indicating a colonoscopy examination site, characterized in that, in step d) above, the image analysis model is composed of an examination site detection model that detects an examination site, a diverticulum detection model that detects a diverticulum site, and a lesion detection model that detects a lesion site.

17. In Paragraph 15, A method for indicating a colonoscopy examination site, characterized in that the above-mentioned lesion detection model is equipped with a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

18. In Paragraph 12, A method for indicating a colonoscopy examination site, characterized in that, in step d) above, the image analysis model is composed of a cecum / diverticulum detection model that detects the cecum and diverticulum, and a lesion detection model that detects the lesion site.

19. In Paragraph 18, A method for indicating a colonoscopy examination site, characterized in that the above-mentioned lesion detection model is equipped with a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

20. In Paragraph 12, A method for indicating a colonoscopy examination site, wherein in step e) above, the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, wherein the examination site includes the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

21. In Paragraph 12, A method for indicating a colonoscopy examination site, characterized in that, in step e) above, the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, and the control unit transmits a command to change the indication condition to the image analysis unit, such that if it is the appendix, the cecum is indicated, and if it is a normal colon image, the diverticulum is indicated differently.

22. In Paragraph 12, A method for indicating a colonoscopy examination site, characterized in that, in step e) above, when the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, the control unit transmits a command to change the indication conditions to the image analysis unit, such as not indicating a diverticulum in a normal colon image, indicating a colonic diverticulum, and indicating the examination time and retrieval time.

23. Model loading / condition setting unit that loads a colonoscopy analysis model and a voice keyword recognition model, and sets the analysis conditions of the analysis model; An image receiver that receives colonoscopy image frames; An image preprocessing unit that preprocesses a colonoscopy image received through the image receiving unit to facilitate subsequent image analysis; An image analysis unit that analyzes a colonoscopy image preprocessed by the image preprocessing unit using an AI (artificial intelligence)-based image analysis model, and detects and indicates at least one of an examination site, a diverticulum site, or a lesion site in the colonoscopy image based on the analyzed result; A voice recognition unit that reads audio from a buffer storing audio while video analysis is being performed by the video analysis unit, analyzes it using an AI-based voice keyword recognition model, recognizes voice keywords based on the analyzed results, and transmits them to the video analysis unit; and A colonoscopy examination site indication system comprising a control unit that controls the status check and operation of the above-mentioned model loading / condition setting unit, image receiving unit, image preprocessing unit, image analysis unit, and voice recognition unit, and when the loading of the colonoscopy image analysis model and the setting of analysis conditions of the analysis model are completed by the above-mentioned model loading / condition setting unit, initializes the analysis screen and displays an image of a normal colon, and provides an analysis result analyzed by the above-mentioned image analysis unit, wherein the analysis result is provided by connecting the analysis target detected by the above-mentioned image analysis model with a voice command (keyword) related to the analysis target uttered by the examiner.

24. In Paragraph 23, Preprocessing of colonoscopy images by the above-mentioned image preprocessing unit is a colonoscopy examination site indication system including analysis area cutting and input size adjustment.

25. In Paragraph 23, A colonoscopy examination site indication system characterized in that the image analysis model of the above-mentioned image analysis unit is composed of a single image analysis model that detects the examination site and the diverticulum site.

26. In Paragraph 23, A colonoscopy examination site indication system characterized in that the image analysis model of the above-mentioned image analysis unit is composed of an examination site detection model that detects the examination site and a diverticulum detection model that detects the diverticulum site.

27. In Paragraph 23, A colonoscopy examination site indication system characterized in that the image analysis model of the image analysis unit is composed of an examination site detection model that detects the examination site, a diverticulum detection model that detects the diverticulum site, and a lesion detection model that detects the lesion site.

28. In Paragraph 27, A colonoscopy examination site indication system characterized by the above-mentioned lesion detection model having a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

29. In Paragraph 23, A colonoscopy examination site indication system characterized by the image analysis model of the above-mentioned image analysis unit being composed of a cecum / diverticulum detection model that detects the cecum and diverticulum and a lesion detection model that detects the lesion site.

30. In Paragraph 29, A colonoscopy examination site indication system characterized by the above-mentioned lesion detection model having a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

31. In Paragraph 23, A colonoscopy examination site indicating system in which the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, wherein the examination site includes the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

32. In Paragraph 23, A colonoscopy examination site indication system characterized by the above-described control unit transmitting an indication condition change command to the above-described image analysis unit, wherein the above-described image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, such that if it is the appendix, the cecum is indicated, and if it is a normal colon image, the diverticulum is indicated differently.

33. In Paragraph 23, A colonoscopy examination site indication system characterized by the above-described control unit transmitting an instruction condition change command to the above-described image analysis unit, wherein the above-described image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, such that the diverticulum is not indicated in the normal colon illustration, the colon diverticulum is indicated, and the examination time and retrieval time are indicated. 34.p) A step in which the model loading / condition setting unit loads the colonoscopy analysis model and the voice keyword recognition model, and sets the analysis conditions of the analysis model; q) A step in which the control unit initializes the analysis screen and displays an image of a normal colon; r) A step in which the image preprocessing unit preprocesses the colonoscopy image received through the image receiving unit so that subsequent image analysis can be performed smoothly; s) A step in which an image analysis unit analyzes a colonoscopy image preprocessed by the image preprocessing unit using an AI-based image analysis model; t) A step in which an image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image based on the analyzed result; u) a step in which a voice recognition unit reads audio from a buffer storing audio while video analysis is being performed by the video analysis unit, analyzes it using an AI-based voice keyword recognition model, recognizes voice keywords based on the analyzed results, and transmits them to the video analysis unit; and v) A method for indicating a colonoscopy examination site comprising the step of a control unit connecting an analysis target detected by an image analysis model of the image analysis unit with a voice command (keyword) related to an analysis target uttered by an examiner to provide an analysis result.

35. In Paragraph 34, A method for indicating a colonoscopy examination site, wherein the preprocessing of the colonoscopy image by the image preprocessing unit in step r) includes cutting the analysis area and adjusting the input size.

36. In Paragraph 34, A method for indicating a colonoscopy examination site, characterized in that, in step s) above, the image analysis model is composed of a single image analysis model that detects the examination site and the diverticulum site.

37. In Paragraph 34, A method for indicating a colonoscopy examination site, characterized in that, in step s) above, the image analysis model is composed of an examination site detection model that detects the examination site and a diverticulum detection model that detects the diverticulum site.

38. In Paragraph 34, A method for indicating a colonoscopy examination site, characterized in that, in step s) above, the image analysis model is composed of an examination site detection model that detects the examination site, a diverticulum detection model that detects the diverticulum site, and a lesion detection model that detects the lesion site.

39. In Paragraph 38, A method for indicating a colonoscopy examination site, characterized in that the above-mentioned lesion detection model is equipped with a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

40. In Paragraph 34, A method for indicating a colonoscopy examination site, characterized in that, in step s) above, the image analysis model is composed of a cecum / diverticulum detection model that detects the cecum and diverticulum, and a lesion detection model that detects the lesion site.

41. In Paragraph 40, A method for indicating a colonoscopy examination site, characterized in that the above-mentioned lesion detection model is equipped with a lesion attribute classification function that distinguishes whether a lesion is benign or malignant.

42. In Paragraph 34, A method for indicating a colonoscopy examination site, wherein in step t) above, the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, wherein the examination site includes the appendix, cecum, ascending colon, transverse colon, descending colon, sigmoid colon, and rectum.

43. In Paragraph 34, A method for indicating a colonoscopy examination site, characterized in that, in step t) above, the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, and the control unit transmits a command to change the indication condition to the image analysis unit, such that if it is the appendix, the cecum is indicated, and if it is a normal colon image, the diverticulum is indicated differently.

44. In Paragraph 34, A method for indicating a colonoscopy examination site, characterized in that, in step t) above, the image analysis unit detects and indicates at least one of an examination site, a diverticulum site, and a lesion site in the colonoscopy image, and the control unit transmits a command to change the indication conditions to the image analysis unit to prevent the diverticulum from being indicated in the normal colon image, to indicate the colon diverticulum, and to indicate the examination time and retrieval time.

Images