Endoscope operation risk early warning system and method based on operation force and image recognition

Abstract

This invention relates to the field of medical endoscopic operation skill assessment technology, and discloses an endoscopic operation risk early warning system and method based on operational force and image recognition. The system includes: an operational force monitoring module that collects multi-directional mechanical signals applied to the endoscope in real time and converts them into four-directional operational force data (push, pull, clockwise and counterclockwise rotation); an endoscopic field of view acquisition module that acquires a video stream; a red screen recognition module that divides the video stream into quadrants, uses an image classification model to identify the red-out characteristics representing the lens contacting the cavity wall, and dynamically calculates the red screen percentage; a central processing unit that simultaneously determines whether the data exceeds a safety threshold, generates a low-level status alert, and non-linearly couples this alert with historical warning frequencies to generate a progressive comprehensive risk alarm level and a real-time quantitative score; and an early warning feedback module that outputs graded early warning signals to the operator and presents the real-time score evolution trend. This invention improves surgical safety and provides an objective digital benchmark for operator skill assessment.

Description

Technical Field

[0001] This invention relates to the field of medical endoscopic operation proficiency assessment technology, specifically to an endoscopic operation risk early warning system and method based on operational force and image recognition. Background Technology

[0002] Endoscopic examination and minimally invasive treatment have become important means of diagnosing and treating diseases of natural cavities. With the increasing number of endoscopic examinations and treatments year by year, the clinical need to ensure surgical safety and reduce intraoperative tissue damage is becoming increasingly urgent.

[0003] Currently, in clinical endoscopic procedures and training, the assessment of endoscopic safety and physician skill levels often relies heavily on the subjective experience of senior physicians. During the procedure, instructors or operators typically rely on visual observation of the screen's field of vision and the physician's hand movements to roughly judge the operational status. This traditional method lacks objective, quantitative data support, making it difficult to standardize the assessment of endoscopic skills and hindering accurate recording and retrospective review of the entire surgical process. Particularly in preventing tissue damage, the lack of an effective tool for objectively, in real-time, and multi-dimensionally monitoring the endoscopic operational status makes it difficult to promptly capture and quantify potentially dangerous operative behaviors in the face of complex and changing cavity environments. This prevents effective intervention before damage occurs, thus necessitating a solution. Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention provides an endoscopic operation risk warning system and method based on operational force and image recognition. It uses objective data fusion to dynamically evaluate the operation process, thereby improving surgical safety and providing an objective digital benchmark for operator skill assessment.

[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention discloses an endoscopic operation risk early warning system based on operating force and image recognition, comprising: The operating force monitoring module is used to collect multi-directional mechanical signals applied by the operator to the endoscope body in real time during endoscopic operations and convert them into four-directional operating force data, including horizontal forward thrust, horizontal backward pull, clockwise rotation force, and counterclockwise rotation force. Endoscopic field of view acquisition module, used to acquire video stream image signals of the endoscopic field of view in real time; The red screen recognition module is connected to the endoscope field of view acquisition module. It is used to divide the acquired video stream image into multiple quadrants according to space, use an image classification model to identify the red screen features in each quadrant that represent the lens touching the cavity wall, and dynamically calculate the red screen percentage data of the red screen in the red screen quadrant in the overall field of view according to a preset time period. The central processing unit is connected to the operation force monitoring module and the red screen recognition module respectively. It is used to synchronously receive the four-way operation force data and the red screen percentage data, and execute nonlinear deep fusion judgment logic: it determines whether the operation force data and red screen percentage data of each direction exceed their respective safety thresholds to generate a low-level status reminder. Then, it nonlinearly couples the real-time status reminders of multi-dimensional parameters with the cumulative warning frequency in the historical operation period to generate a progressive comprehensive risk alarm level and a real-time quantitative risk score. The early warning feedback module, connected to the central processing unit, is used to output a multi-sensory or visual graded early warning signal to the operator based on the dynamic changes of the comprehensive risk alarm level, and simultaneously present the dynamic evolution trend of the real-time quantitative risk score.

[0006] As a further improvement to the above scheme, the red screen recognition module divides a single video stream image into four quadrants; the image classification model classifies the view in each quadrant into one of the following three categories: the out-of-red quadrant view, which represents the field of view being obstructed due to the lens pressing against the cavity wall; the information quadrant view, which represents a clear field of view; and the non-information quadrant view, which represents insufficient cleanliness of the cavity environment.

[0007] As a further improvement to the above scheme, the central processing unit generates low-level status alerts according to the following strategy: When either the horizontal forward thrust or the horizontal backward pull exceeds the set horizontal force threshold, a horizontal force alert is triggered. A rotation force alert is triggered when either the clockwise or counterclockwise rotation force exceeds the set rotation force threshold. Within a preset time period, when the dynamically calculated red screen percentage data exceeds the set red screen percentage threshold, a red screen percentage warning is triggered.

[0008] As a further improvement to the above scheme, the progressive comprehensive risk alarm levels are, in order, gray-card risk alarm, purple-card risk alarm, and black-card risk alarm, and their generation logic is as follows: Triggering either the horizontal force warning or the rotational force warning is defined as a Level 1 operational force warning, while triggering both simultaneously is defined as a Level 2 operational force warning. When either the red screen percentage warning or the level 1 operational force warning is triggered, the gray card risk alarm is output. When both the red screen percentage warning and the first-level operational force warning are triggered simultaneously, or when only the second-level operational force warning is triggered and there is no red screen percentage warning, the purple card risk alarm is output. When the red screen percentage warning and the secondary operational force warning are triggered simultaneously, the black card risk alarm is output.

[0009] As a further improvement to the above scheme, the calculation formula for the real-time quantitative risk score is as follows:

[0010] In the formula, To quantify risk scores in real time, the values ​​range from 0 to 1; The current standardized red-screen ratio; The proportion of the time during which any one of the four-directional maneuvering force data exceeds the safety threshold relative to the actual in-scope time. This is a standardized approach time calculated based on a preset baseline duration; , , These represent the cumulative number of gray, purple, and black card risk alerts triggered during the camera's movement; , , The penalty coefficient is preset based on different risk levels.

[0011] As a further improvement to the above solution, the operating force monitoring module is installed on the control handle of the endoscope, and is equipped with an axial tension / compression sensor and a torque sensor inside, for measuring the multi-directional mechanical signals.

[0012] As a further improvement to the above solution, the early warning feedback module is also used to render a report graph including endoscopic operating force and red screen information in real time and display it through the display interface; the report graph displays the multi-directional operating force numerical distribution curve and the red screen percentage data distributed on different coordinate axis systems, distinguishes horizontal force and rotational force with different visual styles, and superimposes a matching graded risk visual mark at the corresponding time node position when a jump in the comprehensive risk alarm level occurs.

[0013] As a further improvement to the above scheme, the plotting logic of the multi-directional operating force numerical distribution curve is as follows: set the horizontal axis as the time axis and use it as the zero value baseline of the mechanical value; set the horizontal forward thrust and clockwise rotation force as positive values ​​and plot their curves above the zero value baseline; set the horizontal backward pull and counterclockwise rotation force as negative values ​​and plot their curves below the zero value baseline.

[0014] As a further improvement to the above solution, the system is also configured with the following mechanism: Operators can customize or modify the safety thresholds, including the red screen percentage threshold, horizontal force threshold, and rotational force threshold. Before each operation, historical endoscopic operation data is structured and accessed by assigning a unique identification code associated with the endoscope category and personal identification information to the operator, so as to generate a time-series dynamic comparison record of individualized endoscopic operation skills.

[0015] This invention also discloses an endoscopic operation risk warning method based on manipulative force and image recognition, applied to the aforementioned endoscopic operation risk warning system based on manipulative force and image recognition. The warning method includes the following steps: During endoscopic procedures, multi-directional mechanical signals applied by the operator to the endoscope body are collected in real time and converted into four-directional operating force data, including horizontal forward thrust, horizontal backward pull, clockwise rotation force, and counterclockwise rotation force. Real-time acquisition of video stream image signals from the endoscopic field of view; The acquired video stream image is divided into multiple quadrants according to space. The image classification model is used to identify the red-out view features in each quadrant that represent the lens touching the cavity wall. The percentage of red screen data in the red-out quadrant of the overall field of view is dynamically calculated according to a preset time period. The system determines whether the operational force data and red screen percentage data in each direction exceed their respective safety thresholds to generate a basic status alert. Then, it non-linearly couples the real-time status alerts of multi-dimensional parameters with the cumulative warning frequency in the historical operation period to generate a progressive comprehensive risk alarm level and a real-time quantitative risk score. Based on the dynamic changes in the comprehensive risk alarm level, a matching multi-sensory or visual graded early warning signal is output to the operator, and the dynamic evolution trend of the real-time quantitative risk score is presented simultaneously.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention achieves precise real-time quantitative risk scoring and progressive graded alarms by nonlinearly coupling horizontal push-pull and rotational four-dimensional mechanical data during endoscopic operation with red screen percentage data and historical warning frequencies, based on their mutual constraints. This design allows for effective intervention before damage occurs, providing real-time feedback to improve the operator's tip control during endoscopy, reducing medical risks, and ensuring the safety of endoscopic diagnosis and treatment.

[0017] 2. This invention multiplies the percentage of abnormal operational ability, the percentage of red screens, and the time taken for intraoperative observation, ensuring a sharp drop in the total score when any indicator is in a high-risk state. Simultaneously, it uses the historical number of gray, purple, and black card alarms as the exponent of the denominator penalty coefficient, allowing frequent high-risk alarms to have a non-linearly amplified penalty effect. This mechanism realistically reflects the damage a single serious error can cause to overall safety in clinical practice, providing a more objective and scientific digital benchmark for operator skill assessment.

[0018] 3. This invention allows operators to set personalized thresholds based on their own skill level and challenge goals. Based on the assigned associated identification code, it can access and structure data from each endoscopic operation in a structured manner, generating a time-series dynamic comparison record of individual operational skills. This not only helps operators to review and improve themselves, but also provides detailed data support for long-term assessment of departments or teaching and training. Attached Figure Description

[0019] Figure 1 This is a framework diagram of the endoscopic operation risk warning system based on operating force and image recognition in Embodiment 1 of the present invention.

[0020] Figure 2 This is a report diagram including endoscopic operating force and red screen information in Embodiment 1 of the present invention.

[0021] Figure 3 This is a flowchart of the endoscopic operation risk warning method based on operating force and image recognition in Embodiment 2 of the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1

[0024] Please see Figure 1 This embodiment provides an endoscopic operation risk warning system based on operating force and image recognition, including: an operating force monitoring module, an endoscopic field of view acquisition module, a red screen recognition module, a central processing unit, and an early warning feedback module.

[0025] The operating force monitoring module is used to collect multi-directional mechanical signals applied by the operator to the endoscope body in real time during endoscopic operations and convert them into four-directional operating force data, including horizontal forward thrust, horizontal backward pull, clockwise rotation force, and counterclockwise rotation force.

[0026] The operating force monitoring module is installed on the endoscope's control handle or clamped to the endoscope body. It contains axial tension / compression sensors and torque sensors to measure the multi-directional mechanical signals. This module is decoupled from the endoscope device, preserving its original structure. When the operator applies operating force to the endoscope, the elastic body inside the sensor undergoes slight deformation, changing the resistance of the strain gauge and converting it into an analog voltage signal. This signal is then converted into a digital signal by a built-in high-precision analog-to-digital converter. After noise reduction processing such as Kalman filtering, the signal is transmitted in real-time to the central processing unit at a sampling rate of at least 50Hz.

[0027] The endoscopic field of view acquisition module is used to acquire video stream image signals of the endoscopic field of view in real time. Since some endoscopic host devices are confidential or closed, and cannot directly acquire images through the system bus, in this embodiment, the endoscopic field of view acquisition module is preferably a universal external high-definition video stream acquisition card. This acquisition card is bypassed and connected to the output of the endoscopic host, importing the video stream into the early warning system without loss.

[0028] The red screen recognition module is connected to the endoscope field of view acquisition module. It divides the acquired video stream image into multiple quadrants spatially, uses an image classification model to identify the red screen features in each quadrant representing the lens touching the cavity wall, and dynamically calculates the percentage of red screen data in the red-screen quadrants relative to the overall field of view at a preset time period. In this embodiment, the preset time period is preferably 5.0 seconds, meaning that the percentage of red screen data in the previous 5.0 seconds is calculated every 5.0 seconds.

[0029] The red screen recognition module divides a single video stream image into four sector quadrants (i.e., upper left, upper right, lower left, and lower right); the image classification model classifies the view in each quadrant into one of the following three categories: the out-of-red quadrant view, which indicates that the lens is pressed against the cavity wall and obstructs the field of view (in this case, because the lens is close to the mucosal tissue, the field of view is usually red, reflecting improper control of the endoscope tip); the information quadrant view, which indicates that the field of view is clear (the direction of the lumen or mucosal details are visible); and the non-information quadrant view, which indicates that the cleanliness of the cavity environment is insufficient (such as the presence of mucus or feces, which prevents a clear view from being displayed).

[0030] In some embodiments, the image classification model is preferably a convolutional neural network (CNN), such as the ResNet series or MobileNet, to ensure the real-time performance of video stream processing.

[0031] The central processing unit is connected to the operation force monitoring module and the red screen recognition module respectively. It is used to synchronously receive the four-directional operation force data and the red screen percentage data, and execute nonlinear deep fusion judgment logic: it determines whether the operation force data and red screen percentage data in each direction exceed their respective safety thresholds to generate a low-level status reminder. Then, it nonlinearly couples the real-time status reminders of multi-dimensional parameters with the cumulative warning frequency in the historical operation period to generate a progressive comprehensive risk alarm level and a real-time quantitative risk score.

[0032] The central processing unit generates low-level status alerts according to the following strategy: A horizontal force alert is triggered when either the horizontal forward thrust or the horizontal backward pull exceeds a set horizontal force threshold (for example, based on clinical experience in colonoscopy, this horizontal force threshold is preferably set to 9N). When either the clockwise or counterclockwise rotational force exceeds the set rotational force threshold (preferably set to 9N), a rotational force warning is triggered. Within a preset time period, when the dynamically calculated red screen percentage data exceeds the set red screen percentage threshold (for example, this threshold can be set to 40% in a colonoscopy), a red screen percentage warning is triggered.

[0033] The progressive comprehensive risk alarm levels are, in order, gray-card risk alarm, purple-card risk alarm, and black-card risk alarm, and their generation logic is as follows: Triggering either the horizontal force alert or the rotational force alert is defined as a Level 1 operational force warning; triggering both simultaneously is defined as a Level 2 operational force warning. To provide clear auditory guidance in the clinical setting, a buzzer warning of 60 times per minute, 0.1 seconds each time, with a 0.9-second interval, is played 5 times when a Level 1 operational force warning is triggered; a more rapid buzzer of 100 times per minute, 0.1 seconds each time, is played 8 times when a Level 2 operational force warning is triggered. If a Level 2 alarm is triggered during the Level 1 alarm playback, the Level 1 alarm is immediately stopped and the Level 2 alarm is issued.

[0034] When either the red screen percentage warning or the first-level operational force warning is triggered, the gray card risk alarm (representing a single-dimensional operational deviation) is output.

[0035] When both the red screen percentage warning and the first-level operational force warning are triggered simultaneously, or when only the second-level operational force warning is triggered and there is no red screen percentage warning, the purple card risk alarm (representing a moderate damage risk) is output.

[0036] When the red screen percentage warning and the secondary operational force warning are triggered simultaneously, the black plate risk alarm (representing an extremely high risk of damage) is output.

[0037] The formula for calculating the real-time quantitative risk score is as follows:

[0038] In the formula, To quantify risk scores in real time, the values ​​range from 0 to 1; The current standardized red-screen ratio is actually calculated as: actual red-screen ratio / 0.6 × 100%. The proportion of the time during which any one of the four-directional maneuvering force data exceeds the safety threshold relative to the actual in-scope time. This is a standardized approach time calculated based on a preset baseline duration, for example, if the baseline duration is set to 15 minutes. =Actual in-camera time / 15min × 100%; , , These represent the cumulative number of gray, purple, and black card risk alerts triggered during the camera's movement; , , The penalty coefficient is preset based on different risk levels, and is determined according to the clinical risk attenuation weight. , , The optimal values ​​are 1.1, 1.2, and 1.3, respectively. The numerator product structure of this formula ensures that when any one of the parameters is at an extremely high risk, the overall score drops sharply, fully demonstrating the nonlinear coupling constraint relationship between the multidimensional parameters.

[0039] The early warning feedback module is connected to the central processing unit and is used to output a multi-sensory or visual graded early warning signal that matches the dynamic changes of the comprehensive risk alarm level to the operator, and simultaneously present the dynamic evolution trend of the real-time quantitative risk score.

[0040] Please see Figure 2 The early warning feedback module is also used to render a report graph including endoscopic manipulation force and red screen information in real time, and display it through the display interface. The report graph displays the multi-directional manipulation force numerical distribution curve and the red screen percentage data on different coordinate axes, with horizontal force and rotational force distinguished by different visual styles (e.g., horizontal pushing and pulling force is represented by a blue line, and clockwise and counterclockwise rotational force by an orange line). When a jump in the comprehensive risk alarm level occurs, a matching graded risk visual symbol (e.g., an arrow symbol corresponding to a gray, purple, or black card is superimposed below the chart) is superimposed at the corresponding time node. By observing this report graph, the operator can intuitively understand whether the manipulation force at different stages of endoscopy is overloaded and whether the field of vision is maintained as required.

[0041] The plotting logic for the multi-directional operating force numerical distribution curve is as follows: set the horizontal axis as the time axis and use it as the zero value baseline for the mechanical values; set the horizontal forward thrust and clockwise rotation force as positive values ​​and plot their curves above the zero value baseline; set the horizontal backward pull and counterclockwise rotation force as negative values ​​and plot their curves below the zero value baseline.

[0042] In this embodiment, the early warning system is also configured with the following mechanism: Operators can customize or modify the safety thresholds, including red screen percentage threshold, horizontal force threshold, and rotational force threshold (to suit physicians with different existing skill levels and intended challenge targets). Before each operation, historical endoscopic operation data is structured and accessed by assigning a unique identification code associated with the endoscopy category and personal identification information to the operator, so as to generate a time-series dynamic comparison record of individualized endoscopic operation skills (for example, assigning an identification code "Zhang + ID number + colonoscopy" to physician Zhang who performs colonoscopy, so as to facilitate comparison of the dynamic changes in the safety of his endoscopic operations before and after).

[0043] Example 2

[0044] This invention also discloses an endoscopic operation risk warning method based on manipulative force and image recognition, applied to the aforementioned endoscopic operation risk warning system based on manipulative force and image recognition. Please refer to [link / reference]. Figure 3 The early warning method includes the following steps: During endoscopic procedures, multi-directional mechanical signals applied by the operator to the endoscope body are collected in real time and converted into four-directional operating force data, including horizontal forward thrust, horizontal backward pull, clockwise rotation force, and counterclockwise rotation force. Real-time acquisition of video stream image signals from the endoscopic field of view; The acquired video stream image is divided into multiple quadrants according to space. The image classification model is used to identify the red-out view features in each quadrant that represent the lens touching the cavity wall. The percentage of red screen data in the red-out quadrant of the overall field of view is dynamically calculated according to a preset time period. The system determines whether the operational force data and red screen percentage data in each direction exceed their respective safety thresholds to generate a basic status alert. Then, it non-linearly couples the real-time status alerts of multi-dimensional parameters with the cumulative warning frequency in the historical operation period to generate a progressive comprehensive risk alarm level and a real-time quantitative risk score. Based on the dynamic changes in the comprehensive risk alarm level, a matching multi-sensory or visual graded early warning signal is output to the operator, and the dynamic evolution trend of the real-time quantitative risk score is presented simultaneously.

[0045] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An endoscopic operation risk early warning system based on operational force and image recognition, characterized in that, include: The operating force monitoring module is used to collect multi-directional mechanical signals applied by the operator to the endoscope body in real time during endoscopic operations and convert them into four-directional operating force data, including horizontal forward thrust, horizontal backward pull, clockwise rotation force, and counterclockwise rotation force. Endoscopic field of view acquisition module, used to acquire video stream image signals of the endoscopic field of view in real time; The red screen recognition module is connected to the endoscope field of view acquisition module. It is used to divide the acquired video stream image into multiple quadrants according to space, use an image classification model to identify the red screen features in each quadrant that represent the lens touching the cavity wall, and dynamically calculate the red screen percentage data of the red screen in the red screen quadrant in the overall field of view according to a preset time period. The central processing unit is connected to the operation force monitoring module and the red screen recognition module respectively. It is used to synchronously receive the four-way operation force data and the red screen percentage data, and execute nonlinear deep fusion judgment logic: it determines whether the operation force data and red screen percentage data of each direction exceed their respective safety thresholds to generate a low-level status reminder. Then, it nonlinearly couples the real-time status reminders of multi-dimensional parameters with the cumulative warning frequency in the historical operation period to generate a progressive comprehensive risk alarm level and a real-time quantitative risk score. The early warning feedback module, connected to the central processing unit, is used to output a multi-sensory or visual graded early warning signal to the operator based on the dynamic changes of the comprehensive risk alarm level, and simultaneously present the dynamic evolution trend of the real-time quantitative risk score.

2. The endoscopic operation risk warning system based on operating force and image recognition according to claim 1, characterized in that, The red screen recognition module divides a single video stream image into four quadrants; the image classification model classifies the view in each quadrant into one of the following three categories: the out-of-red quadrant view, which represents the view being obstructed by the lens pressing against the cavity wall; the information quadrant view, which represents a clear view; and the non-information quadrant view, which represents insufficient cleanliness of the cavity environment.

3. The endoscopic operation risk early warning system based on operating force and image recognition according to claim 1, characterized in that, The central processing unit generates low-level status alerts according to the following strategy: When either the horizontal forward thrust or the horizontal backward pull exceeds the set horizontal force threshold, a horizontal force alert is triggered. A rotation force alert is triggered when either the clockwise or counterclockwise rotation force exceeds the set rotation force threshold. Within a preset time period, when the dynamically calculated red screen percentage data exceeds the set red screen percentage threshold, a red screen percentage warning is triggered.

4. The endoscopic operation risk early warning system based on operating force and image recognition according to claim 3, characterized in that, The progressive comprehensive risk alarm levels are, in order, gray-card risk alarm, purple-card risk alarm, and black-card risk alarm, and their generation logic is as follows: Triggering either the horizontal force warning or the rotational force warning is defined as a Level 1 operational force warning, while triggering both simultaneously is defined as a Level 2 operational force warning. When either the red screen percentage warning or the level 1 operational force warning is triggered, the gray card risk alarm is output. When both the red screen percentage warning and the first-level operational force warning are triggered simultaneously, or when only the second-level operational force warning is triggered and there is no red screen percentage warning, the purple card risk alarm is output. When the red screen percentage warning and the secondary operational force warning are triggered simultaneously, the black card risk alarm is output.

5. The endoscopic operation risk early warning system based on operating force and image recognition according to claim 4, characterized in that, The formula for calculating the real-time quantitative risk score is as follows: In the formula, To quantify risk scores in real time, the values ​​range from 0 to 1; The current standardized red-screen ratio; The proportion of the time during which any one of the four-directional maneuvering force data exceeds the safety threshold relative to the actual in-scope time. This is a standardized approach time calculated based on a preset baseline duration; , , These represent the cumulative number of gray, purple, and black card risk alerts triggered during the onboarding process, respectively. , , The penalty coefficient is preset based on different risk levels.

6. The endoscopic operation risk early warning system based on operating force and image recognition according to claim 1, characterized in that, The operating force monitoring module is installed on the endoscope's control handle and contains axial tension and compression sensors and torque sensors to measure the multi-directional mechanical signals.

7. The endoscopic operation risk early warning system based on operating force and image recognition according to claim 1, characterized in that, The early warning feedback module is also used to render a report graph including endoscopic operating force and red screen information in real time and display it through the display interface. The report graph displays the multi-directional operating force numerical distribution curve and the red screen percentage data on different coordinate axis systems. Horizontal force and rotational force are distinguished by different visual styles. When a jump in the comprehensive risk alarm level occurs, a matching graded risk visual symbol is superimposed at the corresponding time node position.

8. The endoscopic operation risk early warning system based on operating force and image recognition according to claim 7, characterized in that, The plotting logic for the multi-directional operating force numerical distribution curve is as follows: set the horizontal axis as the time axis and use it as the zero value baseline for the mechanical values; set the horizontal forward thrust and clockwise rotation force as positive values ​​and plot their curves above the zero value baseline; set the horizontal backward pull and counterclockwise rotation force as negative values ​​and plot their curves below the zero value baseline.

9. The endoscopic operation risk early warning system based on operating force and image recognition according to claim 1, characterized in that, The system is also configured with the following mechanisms: Operators can customize or modify the safety thresholds, including the red screen percentage threshold, horizontal force threshold, and rotational force threshold. Before each operation, historical endoscopic operation data is structured and accessed by assigning a unique identification code associated with the endoscope category and personal identification information to the operator, so as to generate a time-series dynamic comparison record of individualized endoscopic operation skills.

10. A method for early warning of endoscopic operation risks based on operational force and image recognition, characterized in that, The endoscopic operation risk warning system based on operating force and image recognition, as described in any one of claims 1 to 9, comprises the following steps: During endoscopic procedures, multi-directional mechanical signals applied by the operator to the endoscope body are collected in real time and converted into four-directional operating force data, including horizontal forward thrust, horizontal backward pull, clockwise rotation force, and counterclockwise rotation force. Real-time acquisition of video stream image signals from the endoscopic field of view; The acquired video stream image is divided into multiple quadrants according to space. The image classification model is used to identify the red-out view features in each quadrant that represent the lens touching the cavity wall. The percentage of red screen data in the red-out quadrant of the overall field of view is dynamically calculated according to a preset time period. The system determines whether the operational force data and red screen percentage data in each direction exceed their respective safety thresholds to generate a basic status alert. Then, it non-linearly couples the real-time status alerts of multi-dimensional parameters with the cumulative warning frequency in the historical operation period to generate a progressive comprehensive risk alarm level and a real-time quantitative risk score. Based on the dynamic changes in the comprehensive risk alarm level, a matching multi-sensory or visual graded early warning signal is output to the operator, and the dynamic evolution trend of the real-time quantitative risk score is presented simultaneously.

Images