Anti-perforation gastrointestinal endoscope device and control method

Abstract

The invention provides an anti-perforation gastrointestinal endoscope device and a control method, and the device comprises a sensing integration module which is arranged at the front end part of an endoscope insertion tube and is used for synchronously obtaining contact pressure data in the circumferential direction of an endoscope body, a microstructure image of a mucous membrane tissue and kinematics data of the endoscope body; an original sensing signal is transmitted to the data processing unit in real time through a signal transmission bus; the data processing unit is in communication connection with the sensing integration module and is used for carrying out space-time alignment on the received multi-mode sensing signals, extracting mucous membrane deformation characteristics and fusing the mucous membrane deformation characteristics to generate a real-time risk score; the data processing unit outputs a multi-level execution instruction corresponding to the risk score; the active intervention module comprises a physical response mechanism distributed at the front end of the endoscope body and a mechanical blocking mechanism distributed at the handle end; the active intervention module receives the execution instruction through the logic control interface and drives a corresponding mechanism to adjust the physical rigidity of the endoscope body or the operation endoscope entering resistance; the punching risk can be effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, and in particular to a perforation-proof gastrointestinal endoscope device and control method. Background Technology

[0002] Traditional endoscopes are commonly used invasive diagnostic and therapeutic instruments in gastroenterology. Their core function is to observe lesions of the digestive tract mucosa through a flexible catheter inserted via the mouth or anus, using a front-end camera and light source. Although modern electronic endoscopes have significantly improved safety, limitations in early instrument design and operating techniques mean that during traditional endoscopy, doctors rely on experience to judge the contact force between the endoscope and tissue, which can easily lead to mucosal damage or perforation due to improper force or complex anatomical structures. Furthermore, current endoscopes lack real-time biomechanical feedback and intelligent warning systems, making it impossible to quantify operational risks.

[0003] Therefore, there is an urgent need for a perforation-proof endoscope device and control method that can effectively reduce the risk of perforation. Summary of the Invention

[0004] The purpose of this invention is to provide a perforation-proof gastrointestinal endoscope device and control method, which aims to solve the technical problem that traditional gastrointestinal endoscopes are prone to mucosal damage or perforation during operation.

[0005] To achieve the above objectives, in a first aspect, the present invention provides a perforation-resistant gastrointestinal endoscope device, comprising: The sensing integration module is located at the front end of the endoscope insertion tube and is used to simultaneously acquire contact pressure data in the circumferential direction of the endoscope, microscopic images of the mucosal tissue, and kinematic data of the endoscope. It transmits the raw sensing signals to the data processing unit in real time through a signal transmission bus built into the insertion tube. The data processing unit, which is communicatively connected to the perception integration module, is used to perform spatiotemporal alignment on the received multimodal sensing signals, extract mucosal deformation features, and fuse them to generate a real-time risk score; the data processing unit outputs multi-level execution instructions corresponding to the risk score through a logic control interface; The active intervention module includes a physical response mechanism distributed at the front end of the scope body and a mechanical blocking mechanism at the handle end; the active intervention module receives the execution command through the logic control interface and drives the corresponding mechanism to adjust the physical stiffness of the scope body or the operating resistance of the scope.

[0006] As a further improvement to the above solution, the perception integration module includes: A ring-shaped pressure sensor array is a flexible array pressure membrane composed of several MEMS pressure sensors arranged in a ring. It covers the peripheral outer surface of the endoscope tip and is used to detect the contact pressure distribution in the circumferential direction of the endoscope. A multispectral imaging ring, coaxially integrated into the endoscope lens, includes a narrowband imaging unit and a tomographic scanning unit; the narrowband imaging unit uses a blue-green dual-band light source to capture the microstructural changes of the mucosal surface and deep tissues. The 3D deformation tracking module is embedded in the sealed chamber at the front end of the endoscope and uses a dual-wavelength laser speckle imaging system to calculate the displacement field of the tissue through optical flow algorithms.

[0007] As a further improvement to the above scheme, the center wavelengths of the blue-green dual-band light source are located in the ranges of 400nm-430nm and 520nm-560nm, respectively; the center wavelengths of the dual-wavelength laser light source are located in the ranges of 700nm-850nm and 1400nm-1600nm, respectively. More preferably, the center wavelengths of the narrowband imaging unit are 415nm and 540nm, respectively; the center wavelength of the tomographic scanning unit is 1310nm; and the center wavelengths of the dual-wavelength laser light source are 780nm and 1520nm, respectively.

[0008] As a further improvement to the above solution, the endoscope tip is also integrated with a negative pressure compensation channel. The negative pressure compensation channel passes through the central region of the multispectral imaging ring in physical space, and its opening and closing state is controlled by the active intervention module according to the risk score, so as to perform local adsorption fixation when the risk of microperforation is detected.

[0009] As a further improvement to the above solution, the sensing integration module is connected to the signal preprocessing bridge inside the endoscope via a multiplexed data bus. The signal preprocessing bridge is used to synchronously transmit the conditioned pressure signal, optical signal and kinematic parameters to the data processing unit.

[0010] As a further improvement to the above scheme, the data processing unit is configured with a mucosal deformation model. The mucosal deformation model adopts the U-Net++ architecture, takes multiple consecutive frames of images as input, detects high-frequency component anomalies caused by muscle fiber rupture through frequency domain analysis, and outputs deformation heatmaps and risk area masks.

[0011] As a further improvement to the above scheme, the data processing unit is configured with a risk decision model. The risk decision model is based on the multi-head attention mechanism of the Transformer architecture and performs weighted fusion of stress features, image features and kinematic features. The weight coefficients of each feature are dynamically adjusted according to the input feature vector containing the patient's age, BMI and intestinal pathology history.

[0012] As a further improvement to the above scheme, the data processing unit is also configured with a cross-modal coupling mechanism to feed back the sensor delay compensation amount output by the Kalman filter to the risk decision model. When data delay is detected, the weight allocation of time-sensitive features is automatically reduced.

[0013] As a further improvement to the above solution, the active intervention module includes: The collapsible snake-bone mechanism, as the physical response mechanism, includes multiple snake-bone segments made of Ni-Ti-Cu shape memory alloy and is disposed in the curved section of the endoscope; the snake-bone segments are connected to a current-controlled actuator at the end of the handle via drive wires, which adjusts the stiffness of the snake-bone through thermal phase change according to the execution command. The emergency braking latch, as the mechanical blocking mechanism, is mechanically coupled to the push rail at the handle end; the emergency braking latch is controlled by an electromagnetic actuator, which is connected to the data processing unit through a control line, and is used to lock the rail when a critical risk is triggered, so that the scope body is in a state where it can only be withdrawn but cannot be pushed forward. The mirror damper, located at the junction of the operating handle and the insertion tube, is used to increase the physical resistance of the mirror in the propulsion direction by adjusting the internal damping circuit.

[0014] As a further improvement to the above solution, the collapsible snake-bone mechanism, emergency braking buckle, and mirror damper are spatially distributed, and in terms of control logic, the data processing unit executes a stepped triggering from soft warning to hard blocking based on the risk score level: When the risk score is low to medium risk, the mirror damper is triggered to provide tactile feedback. When the risk score is high, the current-controlled actuator is triggered to soften the snake mechanism. When the risk score is critical, the electromagnetic actuator is triggered to release the emergency brake latch.

[0015] As a further improvement to the above solution, a flexible drive circuit is embedded inside the collapsible snake-bone mechanism; when the risk score triggers a preset level warning, the drive circuit adjusts the excitation current to reduce the bending stiffness of the shape memory alloy segment from 1200 N / mm. 2 Reduced to 180 N / mm 2 .

[0016] As a further improvement to the above solution, the anti-perforation endoscope device also includes a human-machine interface for overlaying and displaying pressure heatmaps, risk area markers, and vascular density index (VDI) on the display screen.

[0017] Secondly, the present invention also provides a data processing and security control method for the perforation-resistant gastrointestinal endoscope device as described in the first aspect, comprising the following steps executed by a processor:

[0018] It receives in real time the circumferential stress electrical signal of the mirror body, multispectral image pixel data, and mirror body displacement vector transmitted by the sensing integration module; Texture abrupt change features in the image are extracted by frequency domain transformation, and the relative displacement field of the tissue is calculated by optical flow method to quantify mechanical strain rate. The stress electrical signal and strain rate are weighted using the Transformer model to generate a dynamic risk score that reflects the degree of mechanical overload of the mirror body. Based on the rating level, an electrical signal is automatically sent to the actuator to trigger corresponding tactile feedback, propulsion speed limitation, or mirror body physical stiffness switching actions; when the real-time risk rating is critical risk, the electromagnetic actuator is triggered to open the emergency brake latch, physically restricting the mirror entry operation.

[0019] Thirdly, the present invention also provides an electronic terminal, comprising: At least one or more processors; One or more memory units; The memory stores a computer program, and the processor calls the computer program to implement the steps of the data processing and safety control method for the anti-perforation endoscope device as described in the second aspect.

[0020] Because the present invention adopts the above technical solutions, the beneficial effects of this application are as follows: This invention provides a perforation-resistant gastrointestinal endoscope device. Through the deep coupling of multimodal sensing, intelligent decision-making, and active intervention mechanisms, it effectively solves the technical challenge of real-time quantification and effective risk avoidance of mechanical perforation risks in clinical operations. Its beneficial effects are specifically reflected in the following three aspects: 1. Overcoming the limitations of experience-based decision-making, achieving multi-dimensional and precise perception of operational risks. Specifically, through a sensing integration module located at the front end, the traditional single endoscopic image is upgraded into a multi-dimensional data stream including circumferential contact pressure, tissue microstructure, and endoscope kinematics. A ring-shaped pressure sensor array captures the circumferential pressure distribution of the endoscope, combined with a multispectral imaging module to monitor changes in the deep mucosal microstructure, such as muscle fiber status, changing the past practice of doctors relying solely on experience for qualitative mechanical judgments. This fusion of "physical mechanics" and "tissue imaging" sensing methods can quantitatively identify microscopic deformation signs before mechanical perforation, effectively eliminating the blind spots of traditional endoscopes in areas with complex anatomical structures.

[0021] 2. Personalized risk models are constructed to significantly improve the sensitivity and accuracy of early warnings. The data processing unit performs spatiotemporal alignment on the received multimodal signals and extracts mucosal deformation features, generating real-time risk scores through a deep learning model. This not only integrates stress and deformation features but also introduces an adaptive coefficient adjustment mechanism based on patient age, BMI, and pathological history. Weights are allocated through multi-head attention mechanisms using architectures such as Transformer, enabling risk assessment to dynamically evolve according to the intestinal tolerance of different patients. This combination of technologies solves the problems of high false alarm rates and poor scenario specificity caused by fixed parameters in traditional pressure sensing solutions, achieving millisecond-level accurate early warnings and securing a critical window for clinical intervention.

[0022] 3. Achieve closed-loop control of hardware and software, establishing a reliable physical-level safety fuse mechanism. The active intervention module of this invention transforms "passive alerts" into "active risk avoidance" hardware actions by adjusting the stiffness of the endoscope or the operating resistance. When the system determines a high-level risk, the logic control interface drives the front-end physical response mechanism to instantly reduce the bending stiffness of the endoscope, or activates the mechanical blocking mechanism at the handle end, such as a speed-limiting damper or a latch. Through this linkage control from the rear handle to the front end of the endoscope, an automatic fuse-like effect can be generated when mechanical overload is detected. Even if the doctor applies improper force, the device can protect the mucosal tissue through physical vibration reduction, softening, or stopping the insertion of the endoscope, thereby preventing perforation accidents at the source. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0024] Figure 1 This is a block diagram of the overall system architecture of an anti-perforation endoscope device disclosed in this invention; Figure 2 This is a schematic diagram of the structure of the endoscope front-end sensing integrated module disclosed in this invention; Figure 3 This is a schematic diagram showing the annular arrangement of the annular pressure sensor array disclosed in this invention; Figure 4 This is a logical framework diagram of the risk decision-making model disclosed in this invention; Figure 5 This is a flowchart of a data processing and safety control method for an anti-perforation endoscope device disclosed in this invention.

[0025] Figure label: 1. Lens; 2. MEMS pressure sensor.

[0026] The realization of the objective, functional characteristics and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0027] 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 a part of the embodiments of the present invention, and not all of them. 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.

[0028] It should be noted that the technical solutions of the various embodiments of the present invention can be combined with each other, but only if they are based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0029] Example 1

[0030] See Figures 1-4 This embodiment provides a perforation-resistant gastrointestinal endoscope device, designed to address the technical problems of traditional gastrointestinal endoscopy procedures, which rely excessively on physician experience and lack real-time biomechanical feedback, easily leading to mucosal damage or perforation. The perforation-resistant gastrointestinal endoscope device includes a sensing integration module, a data processing unit, and an active intervention module.

[0031] Specifically, the sensing integration module is located at the front end of the endoscope insertion tube and is used to simultaneously acquire circumferential contact pressure data of the endoscope, microscopic images of the mucosal tissue, and kinematic data of the endoscope; the sensing integration module includes the following three parts:

[0032] The ring-shaped pressure sensor array consists of multiple MEMS pressure sensors 2 arranged in a ring, forming a flexible thin film covering the outer surface of the front periphery of the mirror body. In this embodiment, 128 MEMS pressure sensors 2 are arranged in a ring, with a denser arrangement at the front and a sparser arrangement at the back. For details, see [link to documentation]. Figure 3In this embodiment, three sets of MEMS pressure sensor arrays 2 are arranged concentrically and coaxially in a ring around the central axis of the endoscope, and are arranged sequentially from the inside to the outside along the radial direction of the endoscope. These are the inner ring sensor array, the middle ring sensor array, and the outer ring sensor array. The central axes of the three arrays completely coincide with the central axis of the endoscope, ensuring coaxiality and consistency of circumferential force detection. Each MEMS sensor array consists of several independent MEMS pressure sensors 2. All MEMS pressure sensors 2 within the same array are evenly distributed circumferentially, forming a closed ring detection array, achieving full circumferential load uniformity detection. The number of MEMS pressure sensors 2 in the inner, middle, and outer ring sensor arrays increases radially from the inside to the outside, adapting to different circumferential lengths and ensuring that all three arrays can achieve circumferentially evenly spaced arrangement. This design enables continuous detection of the contact pressure distribution along the 360-degree direction of the endoscope, rather than a single point pressure value, allowing for real-time and accurate detection of the circumferential contact pressure distribution between the endoscope and tissue, avoiding mechanical blind spots.

[0033] A multispectral imaging ring, coaxially integrated at the end of the endoscope lens 1, includes a narrowband imaging unit and a tomographic scanning unit. The narrowband imaging unit uses a blue-green dual-band light source to enhance the contrast between blood vessels and structures on the mucosal surface, capturing signs of microscopic lesions or damage. The tomographic scanning unit acquires tissue structural information at a certain depth. Combined, these two units simultaneously capture microstructural changes in both the deep and superficial layers of the mucosa, accurately identifying hidden features preceding mechanical perforation, such as mucosal layer separation or muscle fiber rupture.

[0034] The 3D deformation tracking module, encapsulated within a sealed front chamber, employs a dual-wavelength laser speckle imaging system. By projecting a laser and acquiring the reflected speckle pattern, it uses an optical flow algorithm to calculate the minute displacement field of the mucosal tissue, thereby quantifying the degree and trend of tissue deformation. The specific calculation formula for the optical flow algorithm is shown below: ; in, The wavelength of the laser. For phase difference, To organize the refractive index, It refers to the displacement field or displacement quantity.

[0035] In this embodiment, the center wavelengths of the blue-green dual-band light source are located in the ranges of 400nm-430nm and 520nm-560nm, respectively; the center wavelengths of the dual-wavelength laser light source are located in the ranges of 700nm-850nm and 1400nm-1600nm, respectively; preferably, the center wavelengths of the narrowband imaging unit are 415nm and 540nm, respectively; the center wavelength of the tomographic scanning unit is 1310nm; and the center wavelengths of the dual-wavelength laser light source are 780nm and 1520nm, respectively.

[0036] The raw signals acquired by the three sub-modules are transmitted in real time to the data processing unit through a pre-set signal transmission bus inside the insertion tube.

[0037] The data processing unit, communicatively connected to the sensing integration module, is used to perform spatiotemporal alignment on the received multimodal sensing signals, extract mucosal deformation features, and fuse them to generate a real-time risk score. Specifically, after receiving multimodal data, the data processing unit executes the following processing flow: First, for different sampling rates of pressure, image, and deformation data, the data processing unit uses timestamps for spatiotemporal alignment to ensure the consistency of the analysis benchmark.

[0038] Next, the data processing unit extracts key features, analyzes the peak and gradient changes in the pressure array data, identifies potential pressure concentration points, compares the texture features in the multispectral images to assess the health status of the mucosa, and combines 3D deformation data to calculate the rate and extent of tissue deformation.

[0039] Finally, the data processing unit inputs the extracted features into the risk assessment model, fusing them to generate a real-time risk score. Based on the score results, the data processing unit outputs multi-level execution instructions through the logic control interface; for example, it only records data when the risk is low; outputs a first-level warning instruction when the risk is medium; and outputs a second-level emergency intervention instruction when the risk is high.

[0040] The active intervention module includes a physical response mechanism distributed at the front end of the scope and a mechanical blocking mechanism at the handle end; the active intervention module receives the execution command through the logic control interface and drives the corresponding mechanism to adjust the physical stiffness of the scope or the operating resistance of the scope.

[0041] The active intervention module intervenes in the operation process through two methods according to the instructions of the data processing unit.

[0042] A physical response mechanism, located at the front end of the endoscope, comprises a controlled shape memory alloy or polymer layer. Upon receiving a Level 1 warning command, the mechanism is activated, increasing the local physical stiffness of the endoscope's front end. This increased stiffness alters the endoscope's mechanical properties, transforming its contact with mucosal folds from a bendable "soft contact" to a supportive "hard contact." This effectively disperses local compressive forces, preventing the endoscope from becoming too flexible and embedding itself in the mucosal tissue, causing cuts, while simultaneously ensuring maneuverability when needed.

[0043] The mechanical blocking mechanism, located at the handle end, is typically a resistance device driven by an electromagnetic or electric motor. Upon receiving a secondary emergency intervention command, this mechanism instantly increases the operating resistance or locks the advance / rotation function. This physical force feedback acts directly on the operator's hand, forcing the operation to stop. At the instant the system detects an extremely high risk of perforation, it cuts off the mechanical transmission path of the damaging action, achieving "physical loss prevention."

[0044] During the procedure, the endoscope enters the narrow intestinal segment. The sensing integration module continuously monitors the area. When the contact pressure on one side of the endoscope increases, and the 3D deformation tracking module calculates that the mucosal deformation rate at that point exceeds a threshold, the data processing unit comprehensively determines it as a medium-risk condition and outputs a level-one command. The front-end physical response mechanism is activated, increasing the rigidity of the endoscope's tip. The operator feels the tip harden and can easily rotate the endoscope to pass through, avoiding mucosal dragging damage caused by tip collapse.

[0045] If tissue adhesions are encountered, the pressure array will show multi-point saturation, and the multispectral image will show blurred mucosal texture. The data processing unit will determine this as a high-risk situation and immediately output a secondary command. The mechanical blocking mechanism at the handle end will lock instantly, preventing the operator from advancing further. At this point, with the on-screen warning, the operator will retract the scope and readjust the path, thus avoiding a perforation accident.

[0046] In a preferred embodiment, inside the transparent cap at the very front of the endoscope, in addition to the integrated multispectral imaging ring, a negative pressure compensation channel is also provided. The negative pressure compensation channel is a hollow tubular structure with an inner diameter of one micrometer. Its physical spatial path is designed to pass precisely through the central region of the multispectral imaging ring and connect with the negative pressure source inside the endoscope body.

[0047] This design cleverly integrates the negative pressure channel into the core functional area without affecting the optical field of view of the multispectral imaging ring. The centrally located design ensures that the channel outlet faces the center of the observation field, making the adsorption point highly coincident with the visual monitoring point, thus providing a foundation for subsequent precise local fixation operations.

[0048] The active intervention module adds control functionality to the negative pressure compensation channel, building upon its existing features. After generating a real-time risk score, the data processing unit's logic control interface outputs the aforementioned physical stiffness adjustment or mechanical blocking commands, and also outputs opening and closing commands to the solenoid valve of the negative pressure compensation channel based on the risk characteristics.

[0049] Specifically, when the data processing unit analyzes the data and finds that, although the highest risk level has not been reached, micro-perforation risk characteristics have already appeared, the data processing unit determines it to be in an "very early warning" state. At this time, the logic control interface outputs a command to open the solenoid valve of the negative pressure compensation channel. Within a very short time, under the action of the pressure difference, the channel outlet generates a gentle adsorption force on the local mucosal tissue in the field of view that is in a dangerous state. The adsorption force slightly and temporarily fixes the mucosal tissue around the channel opening, preventing it from moving freely or accumulating with the movement of the endoscope.

[0050] In a preferred embodiment, a signal preprocessing bridge is added between the sensing integration module and the data processing unit, and a multiplexed data bus is used for connection.

[0051] Specifically, the raw signals output by the ring pressure sensor array, multispectral imaging ring, and 3D deformation tracking module integrated within the sensing integration module are first fed into a signal preprocessing bridge located in the middle section of the insertion tube. The signal preprocessing bridge integrates an analog-to-digital converter, a signal amplifier, and a data protocol conversion chip.

[0052] The signal preprocessing bridge performs two main functions. First, it rapidly converts the weak analog signal output from the MEMS pressure sensor 2 into a high-precision digital signal using a dedicated ADC, followed by amplification and noise reduction. Second, it parses and standardizes the data frames for the digital signals from the multispectral imaging ring and the 3D deformation tracking module. This process uniformly modulates the diverse "raw sensing signals" into standardized digital data packets suitable for transmission on the bus, ensuring signal quality and reducing distortion caused by long-distance transmission.

[0053] The signal preprocessing bridge also has a built-in synchronization clock to accurately time all incoming data streams. Subsequently, a multiplexer multiplexes the data streams from all channels onto a single or a few data transmission buses according to preset priorities and timings.

[0054] As a preferred embodiment, to achieve accurate quantification of mucosal morphological changes and identification of muscle fiber breakage risk, this invention configures a mucosal deformation quantification model in the data processing unit. Specifically, the mucosal deformation quantification model built into the data processing unit adopts the U-Net++ architecture. The mucosal deformation quantification model takes continuous multi-frame mucosal tissue images as input, such as dynamic video sequences under endoscopy, and achieves quantitative analysis of mucosal deformation through the following steps: First, the input multi-frame images are preprocessed to extract mucosal motion features between adjacent frames. Then, a frequency domain analysis module is used to perform spectral decomposition on the texture changes of the mucosal tissue, focusing on detecting abnormal high-frequency components generated during muscle fiber rupture. Muscle fiber rupture leads to an increase in local tissue vibration frequency, manifested as a significant increase in the amplitude of high-frequency signals. Finally, based on the multi-scale feature fusion capability of U-Net++, the model combines the frequency domain analysis results with spatial morphological features to output a mucosal deformation heatmap and a risk area mask. The mucosal deformation heatmap visually displays the distribution of deformation degrees in different regions of the mucosa, and the risk area mask marks the anatomical locations with a high risk of muscle fiber rupture.

[0055] In this invention, the U-Net++ architecture enhances the fusion of deep features and shallow details through nested skip connections, enabling it to simultaneously capture macroscopic deformation trends and microscopic texture anomalies of the mucosa, thus solving the problem of insufficient extraction of microscopic muscle fiber breakage features by traditional models. The frequency domain analysis module specifically captures high-frequency signals of muscle fiber breakage, avoiding misjudgments caused by relying solely on spatial domain features and improving the accuracy of risk identification. The output deformation heatmap and risk area mask can be directly used for clinical assessment, helping doctors quickly locate mucosal weak areas and potential breakage risk points, and assisting in the formulation of intervention strategies.

[0056] As a preferred embodiment, to address the problem that existing technologies rely on single risk assessment parameters and cannot accurately assess individual differences, this invention further configures a risk decision-making model in the data processing unit, specifically implemented as follows: The data processing unit incorporates a risk decision-making model based on a multi-head attention mechanism within the Transformer architecture. The risk decision-making model receives three types of input features: first, pressure features, derived from mucosal contact pressure data collected by pressure sensors, such as circumferential pressure gradient, peak pressure location, and pressure change rate; second, image features, obtained through feature extraction from the deformation heatmap output by the mucosal deformation model and the risk area mask, such as deformation area, maximum strain rate, and high-frequency component energy value; and third, kinematic features, reflecting dynamic parameters such as displacement and velocity of the mucosa during intestinal peristalsis, such as the instantaneous velocity of the endoscope and rotational angular acceleration. The model maps these three types of features to a unified feature space through the multi-head attention mechanism and performs weighted fusion to form a comprehensive risk representation vector.

[0057] Building upon this foundation, the model incorporates patient-specific feature vectors as the basis for weight adjustment. These vectors consist of the patient's age, body mass index (BMI), and past intestinal pathology history (such as previous inflammation, ulceration, or surgical history). The model internally employs a learnable weight mapping function to dynamically adjust the weight coefficients of pressure features, image features, and kinematic features during the fusion process based on the input individualized feature vectors.

[0058] Specifically, the method for dynamically and adaptively allocating risk weights using individualized patient feature vectors is as follows: Patient-specific feature vector construction: First, the patient's individual characteristics are encoded into a one-dimensional feature vector. ,in, The intestinal pathology history score is used to characterize the risk of underlying intestinal diseases. It is a dimensionless number ranging from 0 to 1. For example, 0 indicates no history of disease, and 1 indicates a severe pathological condition such as ulcers or diverticula. Representing age, BMI stands for Body Mass Index, used to characterize a patient's body shape, and is calculated as weight / height.

[0059] Dynamic adaptive weight generation based on individualized features: By utilizing a built-in lightweight fully connected network and optimizing the Softmax output layer, the system adapts the input patient feature vector. The normalized weight coefficients for the three risk factors are dynamically generated, and the calculation formula is as follows: ; The above formula naturally satisfies the constraints. ; For risk factor category index, These correspond to three core risk factors: pressure gradient risk, maximum stress area risk, and intestinal volume change rate risk, respectively. For the first The dynamic adaptive weight coefficients corresponding to risk factors are dimensionless numbers used to characterize the contribution ratio of the corresponding risk factors in the comprehensive score. For trainable weight row vectors, the dimensions and feature vectors are given. Matching is used to establish a mapping relationship between individual patient characteristics and corresponding risk factors. This is a trainable bias term used to adjust the baseline safety margin for the corresponding risk category, and it is a dimensionless number. and All datasets were pre-trained using a clinical risk dataset for intestinal diagnosis and treatment, with adverse events such as intestinal perforation and bleeding as supervised labels to ensure the clinical rationality of the weight mapping.

[0060] Definition and dimensionless preprocessing of risk factors: To eliminate the differences in physical dimensions among different risk factors and avoid scoring distortion caused by dimensional differences, the three types of original risk factors are first preprocessed by dimensionless conversion and interval mapping. The preprocessing formula is as follows: ; in For the k-th type of original risk factor, there are 3 types of core risk indicators, specifically defined as follows: This represents the pressure gradient in the intestinal lumen during intestinal manipulation, characterizing the severity of local pressure changes in the intestinal wall, and is expressed in Pa / m. This represents the area of ​​maximum stress on the intestinal wall, characterizing the extent of high-stress regions in the intestinal wall, i.e., the cumulative area where intestinal wall stress exceeds the clinical safety threshold, expressed in mm. 2 ; It represents the instantaneous rate of change of intestinal lumen volume, characterizing the rate of expansion / contraction of the intestinal lumen, with units of mL / s; Here, represents the feature scaling factor corresponding to the k-th risk factor, and is a pre-calibrated dimensionless constant. Its value is determined based on the clinical safety threshold of the corresponding risk factor and the numerical distribution of a large clinical sample. Its core function is to scale the original risk factors... Mapped to the sensitive region [-5,5] of the Sigmoid activation function, while simultaneously performing dimensionless processing; The dimensionless eigenvalues ​​of the k-th type of risk factor after preprocessing.

[0061] Individualized comprehensive risk score calculation: After completing the adaptive weight generation and risk factor preprocessing, the normalized individualized comprehensive risk score is calculated using the following formula: ; The Sigmoid activation function maps the preprocessed risk features to the risk probability interval (0,1), and its function expression is: .

[0062] The final result This is a dimensionless number ranging from (0,1), with values ​​closer to 1 indicating a higher overall risk for the patient's current intestinal procedure. This method utilizes dynamic adaptive weights. Instead of traditional fixed weights, the contribution ratio of different risk factors can be adjusted in real time according to the patient's age, BMI, and intestinal pathology history. This solves the problem that fixed weights cannot adapt to individual differences among patients, making the risk score more personalized and clinically valuable.

[0063] The aforementioned cross-modal coupling design can dynamically adjust the weighting coefficients based on age. Since elderly patients have thinner intestinal walls and are more sensitive to compressive stress, the system automatically increases the weighting coefficient of pressure characteristics. When based on body shape, for obese patients with complex and distorted intestinal anatomy, the system automatically increases the weighting coefficient of the maximum stress area. When based on pathology, for patients with a history of inflammation or ulcers, the system automatically increases the weighting of the rate of volume change to sensitively detect the risk of microperforation caused by stress-induced expansion.

[0064] In a preferred embodiment, in a multimodal sensing system, the sampling frequencies and hardware transmission links for different physical quantities (such as pressure, images, and kinematics) often differ. If data transmission delays occur in complex clinical environments, directly using data with time differences for fusion calculations, especially when extracting features requiring differential operations such as "pressure change rate," will generate significant calculation errors. This can lead to frequent false alarms in the anti-perforation system, interfering with the doctor's normal operation. To address this technical problem caused by asynchronous multi-source data, this embodiment further optimizes the data processing unit, as detailed below: The data processing unit is internally configured with a cross-modal coupling mechanism. Specifically, the system front-end first performs spatiotemporal alignment processing on the pressure and image sensing data using Kalman filtering to synchronize timestamps and compensate for sensor delays. During this spatiotemporal alignment operation, the Kalman filter calculates and outputs a sensing delay compensation amount in real time, denoted as [missing value]. The system does not simply send the passively aligned data to the next level algorithm; instead, it constructs a closed-loop control loop of "sensor delay - risk weight." The system uses the sensor delay compensation amount output from the Kalman filter as... As a dynamic adjustment factor, it is directly fed back into the risk decision-making model, that is, injected into the weight network used to generate adaptive coefficients. The specific formula for adjusting the weight coefficients is shown below: ; in, It is a learnable model parameter, automatically optimized by the neural network during training, used to dynamically adjust the sensing delay compensation amount Δt on the weights. The intensity of the impact.

[0065] When the system detects a delay in data from a certain sensor in real time, that is... When the pressure is greater than 0, the cross-modal coupling mechanism intervenes, automatically reducing the model's dependence on time-sensitive features, such as lowering the weight allocation of factors like "pressure change rate". Time-sensitive features typically include derivative calculations over time, which require extremely high timestamp accuracy. Even minute data delays, if directly substituted into the calculation without processing, will be amplified at the algorithm level, triggering system misjudgments, such as mistakenly believing the endoscope is being rapidly and violently advanced. This invention, through the cross-modal coupling mechanism, proactively weakens the decision-making power of time-sensitive features at the physical instant of data delay, instead allowing the system to refer more to non-time-sensitive static features such as absolute pressure peaks or image deformation area. This feature combination effectively avoids false high-risk alarms caused by complex electromagnetic interference, occasional hardware glitches, or transmission delays in the medical setting, significantly improving the robustness and decision reliability of the anti-perforation warning system in real clinical conditions.

[0066] The following example, using a patient as an example, further illustrates the inventive concept of this invention. The patient is 65 years old, has a BMI of 28, and a history of colonic ulcers, that is... .

[0067] Before the endoscope was inserted, the model automatically calculated the feature weights assigned to the patient as follows: , , Because the patient was elderly and had a history of peptic ulcer disease, age and pathological history dominated the weighting, causing the system to be in a state of highest sensitivity and alertness to stress characteristics.

[0068] In practical operation, when the sensor array detects the rate of local pressure change in real time... At pressures exceeding 15 kPa / s, the system will bypass the conventional deformation accumulation assessment process and trigger a high-level over-pressure alarm immediately, as the pressure coefficient becomes the dominant factor. Subsequently, the active intervention module will immediately intervene to soften the endoscope or release the latches.

[0069] As can be seen from the specific algorithm implementations described above, this device completely abandons the "one-size-fits-all" fixed alarm threshold. Based on the patient's actual anatomical tolerance limits, it can achieve precise quantitative protection tailored to each individual within a millisecond-level time window, effectively blocking the continuous output of destructive stress from its physical source before mechanical perforation actually occurs.

[0070] As a preferred embodiment, to address the technical problem that traditional gastrointestinal endoscopes lack physical-level active perforation protection and rely entirely on the doctor's subjective operation, making them highly susceptible to tissue damage in unexpected situations, this embodiment further designs multi-level protection for the mechanical structure and control logic of the active intervention module. Specifically, the active intervention module includes: A collapsible snake-bone mechanism, as a physical response mechanism distributed at the front end, is located in the curved section of the endoscope and includes multiple snake-bone segments made of Ni-Ti-Cu shape memory alloy. These snake-bone segments are electrically connected to a current-controlled actuator at the handle end via built-in drive wires. Based on real-time execution commands, the system adjusts the output current through the current-controlled actuator, dynamically adjusting the bending stiffness of the snake-bone using the thermally induced phase change characteristics of the shape memory alloy. Traditional endoscopes have relatively fixed stiffness in the curved section at the front end. When encountering sharp angles or physiological narrowings in the intestine, the endoscope is prone to puncturing the intestinal wall due to excessive stiffness and uneven stress. In this embodiment, when the system detects a level 4 critical risk, the actuator causes an instantaneous phase change in the alloy, resulting in the endoscope collapsing. The bending stiffness of the endoscope can be significantly reduced, for example, from 1200 N / mm. 2 Reduced to 180 N / mm 2 The core of this design lies in the fact that it is equivalent to installing a "mechanical fuse" at the very front of the device; at the moment when the mechanical tolerance limit of the mucosal tissue is exceeded, the front end of the endoscope actively softens, and at the cost of losing its own propulsion and force transmission capabilities, it instantly unloads the destructive concentrated stress applied to the tube wall, thereby preventing perforation from the physical source.

[0071] A scope damper is installed at the junction of the operating handle and the insertion tube. During scope insertion, when the data processing unit determines the risk score to be medium to high, the system automatically intervenes and adjusts the damping circuit inside the scope damper, thereby increasing the physical resistance of the scope in the advancement direction. With traditional equipment, the resistance during scope insertion mainly comes from the natural friction of the tube wall, and doctors often cannot perceive subtle changes in resistance. The scope damper physically limits the instantaneous advancement speed of the scope. It directly converts the abstract risk score calculated by the system into a reverse physical resistance that the operator can intuitively perceive with their hand. This not only provides mandatory tactile feedback warnings but also kinetically limits the operator's unconscious, forceful advancement movements, providing crucial buffer time for subsequent diagnostic decisions or angle adjustments.

[0072] The emergency braking mechanism, located at the handle end, is a mechanical stop mechanism mechanically coupled to the guide rail and controlled by an electromagnetic actuator connected to the data processing unit's control circuitry. When the system determines that the operation is extremely dangerous, such as triggering a critical risk threshold, the electromagnetic actuator instantly activates, causing the latch to pop open and directly lock the guide rail. At this point, the scope's physical movement is forcibly changed to a one-way locked state of "only retreating, not advancing." In extremely high-risk situations or when the operator is under stress, simply softening the front stiffness or increasing damping may not be enough to prevent a few doctors from applying inertial force. This feature provides the highest level of physical hard-limit protection at the handle control end. When a critical risk occurs, the system forcibly cuts off the operator's scope advancement permission, eliminating the possibility of secondary injury due to subjective misjudgment; at the same time, retaining the scope retreat permission ensures the safe removal of the equipment from the human body. This design, from soft warning to hard stripping of advancement permission, completely eliminates the possibility of perforation due to uncontrollable human intervention.

[0073] As a preferred embodiment, to address the technical problems of traditional equipment intervention methods being singular and abrupt interventions easily disrupting normal diagnostic and treatment processes, this embodiment further defines the systematic architecture of the active intervention module in terms of space and control logic. Specifically, the collapsible snake-bone mechanism, emergency braking latch, and endoscope damper are spatially distributed.

[0074] Specifically, the collapsible snake-bone mechanism is deployed at the curved section of the endoscope's tip, which directly contacts the tissue, while the endoscope damper and emergency brake buckle are located at the end of the operating handle where the doctor applies force. This distributed physical layout of "flexible stress relief at the front and rigid limiting at the rear" can both eliminate destructive stress directly at the tissue end and cut off the power source at the operating end, thereby constructing a complete front-to-back coordinated protective barrier along the entire instrument pathway.

[0075] In terms of control logic, the data processing unit does not employ a single fixed threshold alarm. Instead, it executes a tiered triggering strategy, from soft warning to hard blocking, based on real-time quantified risk scores. For different risk levels, the system specifically performs the following response actions:

[0076] The soft-touch warning stage, which is also the low-to-medium risk stage: When the data processing unit determines that the current operation falls within the low-to-medium risk range, the system first triggers the endoscope damper at the handle, increasing the propulsion resistance by adjusting the internal damping circuit. In low-to-medium risk conditions, such as when the endoscope just touches the intestinal wall and undergoes slight deformation, the system does not forcibly deprive the doctor of their operational authority. Instead, it converts the digitized risk score into physical resistance that the doctor can intuitively perceive with their hand. This dynamic feedback of tactile feedback provides timely and gentle reminders, giving the doctor a buffer window to proactively adjust the angle or force of the endoscope insertion. This ensures safety while maximizing the smoothness of clinical examinations, avoiding frequent mechanical interventions that could interrupt normal diagnostic and treatment procedures.

[0077] The initial physical stress release phase, which is also a high-risk phase: When the risk score rises to a high-risk level, such as when the tissue deformation rate approaches a critical value, the system triggers a current-controlled actuator to output an excitation current to the front-end serpentine mechanism, causing it to undergo a thermal phase transition and soften instantaneously. In traditional rigid endoscopes, when encountering high resistance conditions, the doctor's continuous thrust directly translates into destructive force that punctures the intestinal wall. This embodiment, through the instantaneous collapse of the front-end serpentine stiffness, eliminates the physical basis of "rigid force transmission" at the endoscope's front end. At this point, even if the doctor unconsciously applies thrust at the rear end, the thrust will be absorbed by the softened front end itself and cannot be effectively transmitted to the vessel wall, thereby unloading mechanical overload stress at the microscopic tissue level and achieving active physical risk avoidance.

[0078] The final stage of mechanical hard blockade, which is also the critical risk stage: When the risk score reaches a critical level, such as when OCT detects an impending muscle fiber rupture or extreme local pressure, the system triggers the electromagnetic actuator at the handle end, causing the emergency brake latch to instantly pop open and directly lock the advance guide rail. In extreme critical situations, a single damping indicator or front-end softening may not be sufficient to counteract the doctor's inertial force or the patient's sudden, severe spasm. At this point, the popping of the brake latch forcibly cuts off the advance transmission link at the mechanical control source, physically locking the scope into a "one-way state where it can only be withdrawn but not advanced." This mandatory hard-blocking mechanism completely eliminates the possibility of perforation caused by uncontrollable factors such as operator misjudgment or stress response, constructing the final, absolutely safe physical safety barrier for the entire perforation prevention system.

[0079] As a preferred embodiment, in order to solve the technical problem that when a traditional endoscope is advanced in a narrow or sharply curved intestine, the non-adjustable stiffness of the endoscope tip can easily lead to local tissue stress concentration or even direct perforation of the intestinal wall, this embodiment further discloses the specific circuit response and mechanical collapse characteristics of the collapsible snake bone mechanism.

[0080] The collapsible snake-bone mechanism incorporates a flexible drive circuit. This flexible drive circuit is routed along the curved sections of the endoscope, perfectly adapting to the complex bending deformations of the endoscope within the body without increasing the overall outer diameter. When the risk score calculated by the data processing unit triggers a preset warning level, such as when the system determines a high risk of perforation, the flexible drive circuit rapidly receives and responds to the instruction. This drive circuit precisely adjusts the output excitation current to instantaneously heat the snake-bone segments made of shape memory alloy, inducing a thermally induced phase transition. Under this phase transition, the bending stiffness of the shape memory alloy segments increases from 1200 N / mm under normal conditions. 2 The pressure dropped sharply to 180 N / mm 2 .

[0081] Under normal safety conditions, 1200 N / mm 2 The bending stiffness is sufficient to ensure that the endoscope has good guidance and force transmission performance, allowing doctors to smoothly and steadily complete intestinal crossing operations. When encountering anatomical variations, sclerosis of lesions, or improper force application by the doctor, causing local contact stress to approach the critical point of tissue rupture, the system triggers a precipitous drop in stiffness. This rapid softening process instantly transforms the tip of the endoscope from a "rigid probe" into a "flexible, collapsible body." At this point, even if the operator continues to apply a large pushing force to the handle due to inertia or insufficient reaction, this mechanical pushing force will be completely absorbed by the bending deformation of the extremely softened tip snake bone itself, resulting in a significant reduction in the actual operating force applied to the tip mucosa. This physical active force relief mechanism completely blocks the transmission of the rear pushing force to the front damaging force, providing absolute physical safety for the fragile intestinal mucosa without completely interrupting the doctor's operation.

[0082] As a preferred embodiment, in order to solve the technical problem that doctors can only observe the surface image of the tissue in traditional gastrointestinal endoscopy operations and cannot intuitively perceive the force distribution and micro-compression state between the endoscope and the tissue, which can easily lead to perforation due to blind force in complex lumens, this embodiment further optimizes the feedback mechanism of the system and discloses a specific embodiment of the human-computer interaction interface.

[0083] In this embodiment, the anti-perforation endoscope device also includes a human-machine interface. The human-machine interface is communicatively connected to the data processing unit and is used to simultaneously overlay and display pressure heatmaps, risk area markers, and vascular density index (VDI) on the display screen based on conventional endoscopic video images using AR visualization technology.

[0084] The overlay display of pressure heatmaps reveals that in traditional procedures, doctors rely entirely on manual perception to judge contact force, resulting in significant subjectivity and latency. This device extracts data from a ring-shaped pressure sensor array, renders it as a color heatmap that changes with stress gradients, and attaches it in real-time to the endoscopic video feed. This feature transforms invisible physical contact stress into intuitive color signals visible to the operator. While searching for lesions on the screen, doctors can precisely "see" the stress concentration points between the endoscope and the intestinal wall, allowing them to consciously adjust the insertion angle and operating path, avoiding destructive compression of the local intestinal wall from the outset.

[0085] Risk Area Marking: Subtle tissue deformations within the complex anatomy of the intestine are extremely difficult to observe directly. This interface highlights or borders high-risk masked areas identified by the algorithm directly on the screen. It overcomes the physiological limitations of human visual perception, directly presenting doctors with the microscopic anomalies captured by the underlying algorithm through frequency domain analysis and optical flow. This is equivalent to equipping doctors with "mechanical X-ray vision," exposing tissue mechanical weaknesses before macroscopic physical perforation occurs, helping doctors avoid high-risk areas.

[0086] The quantification of the Vascular Density Index (VDI) is as follows: VDI is a quantitative indicator calculated based on multispectral imaging data. When the intestinal wall becomes thinner due to excessive traction or compression by the endoscope, the submucosal capillaries will contract under pressure, leading to a decrease in the VDI value. Specifically, the calculation model for the Vascular Density Index is shown below: ; in, Let i be the area of ​​the i-th blood vessel region. For the area of ​​the region of interest, wavelength The light intensity below, The light intensity at a wavelength of 540nm. is the light intensity at a wavelength of 415nm, and n is the total number of blood vessel regions identified within the current region of interest.

[0087] The real-time display of VDI values ​​or change curves on the monitor provides doctors with a highly valuable clinical indicator of "tissue compression and ischemia." It can provide earlier warnings of deformation overload through changes in blood perfusion before substantial tearing of the tissue structure occurs, further improving the lead time and sensitivity of perforation prevention. In summary, this human-computer interface, through the simultaneous display of multi-dimensional information, completely changes the traditional "blind men and the elephant" tentative operation, constructing a new paradigm of diagnosis and treatment from "single visual observation" to a multi-dimensional perspective of "mechanics-structure-physiology," significantly reducing the probability of accidental injuries caused by visual blind spots and dulled tactile perception.

[0088] Example 2

[0089] To address the technical problem of traditional gastrointestinal endoscopy procedures relying too heavily on physicians' subjective experience and lacking objective biomechanical quantification methods, thus easily leading to perforation, see [reference needed]. Figure 5 This embodiment also provides a data processing and safety control method for an anti-perforation gastrointestinal endoscope device. This method is automatically executed by the data processing unit (processor) inside the device, and its essence is the low-level control logic of the physical operating state of the medical device. The specific implementation steps are as follows: S1. Synchronous acquisition of multidimensional physical signals: The processor receives in real time the circumferential stress electrical signal of the mirror body, multispectral image pixel data, and mirror body displacement vector transmitted by the sensing integration module.

[0090] This step forms the data foundation of the entire system. By converting the physical interaction of the endoscope within the digestive tract into objective electrical signals and pixel vectors, the system completely eliminates the blind spot of "guessing by feel" in traditional operations. The synchronous input of multi-source data ensures that subsequent control logic can be deduced based on the most complete and realistic environmental mechanical feedback.

[0091] S2. Quantitative Extraction of Mechanical Deformation Features: The processor extracts texture abrupt change features from the multispectral image through frequency domain transformation and calculates the relative displacement field of the tissue using optical flow, thereby quantifying the mechanical strain rate of the tissue; preferably, a method such as Fast Fourier Transform (FFT) is used to extract texture abrupt change features from the multispectral image. The specific calculation formula of the optical flow algorithm is shown below: ; in, The wavelength of the laser. For phase difference, To organize the refractive index, It refers to the displacement field or displacement quantity.

[0092] During endoscopic advancement, before macroscopic rupture occurs, the intestinal wall tissue first exhibits stretching, separation, or breakage of its microscopic muscle fibers. These microscopic changes are difficult to detect with the naked eye in conventional video, but they manifest as significant high-frequency component anomalies in frequency domain analysis. By combining optical flow methods to dynamically track the tissue displacement field, this step can capture abnormal stretching states in the submucosa with extremely high sensitivity. This significantly advances the system's risk warning window, achieving a leap from "post-event remediation" to "early intervention in microscopic deformation."

[0093] S3. Calculation of dynamic risk score: The processor uses the Transformer model to weight the input stress electrical signal and mechanical strain rate to generate a dynamic risk score that reflects the degree of mechanical overload of the mirror body.

[0094] Traditional perforation prevention solutions often employ rigid, fixed thresholds, triggering an alarm when pressure exceeds a certain value. This is highly susceptible to frequent false alarms due to individual patient differences. This step introduces the Transformer's multi-head attention mechanism, which dynamically focuses on the most critical risk factors based on real-time conditions. For example, at a given moment, the weight of a sudden stress surge is adaptively amplified. The dynamic risk score objectively reflects the degree of mechanical overload currently applied to the cannula by the endoscope, rather than drawing a medical diagnosis. This ensures accurate assessment of mechanical risks while effectively improving system robustness under complex clinical conditions.

[0095] S4. Hardware-level stepped response and physical limitations: The processor automatically sends a level signal to the back-end actuator based on the level of the dynamic risk score to trigger corresponding tactile feedback, propulsion speed limitation or scope physical stiffness switching actions; in particular, when the real-time risk score is critical risk, the electromagnetic actuator is triggered to open the emergency brake latch, physically limiting the scope entry operation.

[0096] This is a crucial step in forming a safe closed loop for this control method. The system transforms abstract risk scores into substantial mechanical resistance and physical limitations. In low- and medium-risk situations, the system uses tactile feedback and damping speed limiting to provide gentle reminders to the physician while preserving their operational control and preventing interruptions to normal examinations. In critical situations, such as when mechanical overload is about to exceed the tissue's physical tolerance limit, the system bypasses manual operation and directly drives the electromagnetic actuator to release the brake latch, forcibly locking the scope in a unidirectional "reverse-only" state. This mechanism constructs the final physical safety barrier, unaffected by human error or fatigue, completely cutting off the injury-causing kinetic energy leading to perforation at the source of mechanical control.

[0097] Example 3

[0098] The present invention also provides an electronic terminal, comprising: A memory on which computer programs are stored; A processor is used to load and execute the computer program to implement the data processing and safety control method for a perforation-resistant gastrointestinal endoscope device as described in Embodiment 2.

[0099] A processor may include one or more processing units, such as an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural network processing unit (NPU). These different processing units may be independent devices or integrated into one or more processors.

[0100] The controller can serve as the nerve center and command center of an electronic device. Based on the instruction opcode and timing signals, the controller generates operation control signals to control the fetching and execution of instructions.

[0101] The processor may also include memory for storing instructions and data. In some embodiments, the memory in the processor is a cache memory. This memory can store instructions or data that the processor has just used or that are used repeatedly. If the processor needs to use the instruction or data again, it can retrieve it directly from the memory. This avoids repeated accesses, reduces processor waiting time, and thus improves system efficiency.

[0102] The above are merely preferred embodiments of the present invention and do not limit the patent scope of the present invention. Any equivalent structural transformations made using the contents of the present invention's specification and drawings under the inventive concept of the present invention, or direct or indirect applications in other related technical fields, are within the patent protection scope of the present invention.

Claims

1. A perforation-resistant gastrointestinal endoscope device, characterized in that, include: The sensing integration module is located at the front end of the endoscope insertion tube and is used to simultaneously acquire contact pressure data in the circumferential direction of the endoscope, microscopic images of the mucosal tissue, and kinematic data of the endoscope. It transmits the raw sensing signal to the data processing unit in real time through a signal transmission bus built into the insertion tube; The data processing unit, which is communicatively connected to the perception integration module, is used to perform spatiotemporal alignment on the received multimodal sensing signals, extract mucosal deformation features, and fuse them to generate a real-time risk score. The data processing unit outputs multi-level execution instructions corresponding to the risk score through a logic control interface; The active intervention module includes a physical response mechanism distributed at the front end of the scope body and a mechanical blocking mechanism at the handle end; the active intervention module receives the execution command through the logic control interface and drives the corresponding mechanism to adjust the physical stiffness of the scope body or the operating resistance of the scope.

2. The anti-perforation endoscope device according to claim 1, characterized in that, The perception integration module includes: A ring-shaped pressure sensor array consists of a flexible array pressure membrane composed of several MEMS pressure sensors arranged in a ring, which covers the peripheral outer surface of the endoscope tip and is used to detect the contact pressure distribution in the circumferential direction of the endoscope. A multispectral imaging ring, coaxially integrated into the endoscope lens, includes a narrowband imaging unit and a tomographic scanning unit; the narrowband imaging unit uses a blue-green dual-band light source to capture the microstructural changes of the mucosal surface and deep tissues. The 3D deformation tracking module is embedded in the sealed chamber at the front end of the endoscope and uses a dual-wavelength laser speckle imaging system to calculate the displacement field of the tissue through optical flow algorithms.

3. The anti-perforation endoscope device according to claim 1, characterized in that, The data processing unit is equipped with a mucosal deformation model. The mucosal deformation model adopts the U-Net++ architecture, takes multiple consecutive frames of images as input, detects high-frequency component anomalies caused by muscle fiber rupture through frequency domain analysis, and outputs deformation heatmaps and risk area masks.

4. The anti-perforation endoscope device according to claim 1, characterized in that, The data processing unit is equipped with a risk decision model, which is based on the multi-head attention mechanism of the Transformer architecture and weightedly fuses stress features, image features and kinematic features. The weight coefficients of each feature are dynamically adjusted based on the input feature vector containing the patient's age, BMI, and intestinal pathology history.

5. The anti-perforation endoscope device according to claim 4, characterized in that, The data processing unit is also equipped with a cross-modal coupling mechanism to feed back the sensor delay compensation amount output by the Kalman filter to the risk decision model. When data delay is detected, the weight allocation of time-sensitive features is automatically reduced.

6. The anti-perforation endoscope device according to claim 1, characterized in that, The active intervention module includes: The collapsible snake-bone mechanism, as the physical response mechanism, includes multiple snake-bone segments made of Ni-Ti-Cu shape memory alloy and is disposed in the curved section of the endoscope; the snake-bone segments are connected to a current-controlled actuator at the end of the handle via drive wires, which adjusts the stiffness of the snake-bone through thermal phase change according to the execution command. The emergency braking latch, as the mechanical blocking mechanism, is mechanically coupled to the push rail at the handle end; the emergency braking latch is controlled by an electromagnetic actuator, which is connected to the data processing unit through a control line, and is used to lock the rail when a critical risk is triggered, so that the scope body is in a state where it can only be withdrawn but cannot be pushed forward. The mirror damper, located at the junction of the operating handle and the insertion tube, is used to increase the physical resistance of the mirror in the propulsion direction by adjusting the internal damping circuit.

7. The anti-perforation endoscope device according to claim 6, characterized in that, The collapsible snake-bone mechanism, emergency braking latch, and mirror damper are spatially distributed. In terms of control logic, the data processing unit executes a stepped triggering process from soft warning to hard blocking based on the risk score level. When the risk score is low to medium risk, the mirror damper is triggered to provide tactile feedback. When the risk score is high, the current-controlled actuator is triggered to soften the snake mechanism. When the risk score is critical, the electromagnetic actuator is triggered to release the emergency brake latch.

8. The anti-perforation endoscope device according to claim 7, characterized in that, The collapsible serpentine mechanism incorporates a flexible drive circuit; when the risk score triggers a preset level warning, the drive circuit adjusts the excitation current to reduce the bending stiffness of the shape memory alloy segment from 1200 N / mm. 2 Reduced to 180 N / mm 2 .

9. The anti-perforation endoscope device according to claim 1, characterized in that, The perforation-resistant gastrointestinal endoscope device also includes a human-machine interface for overlaying and displaying pressure heat maps, risk area markers, and vascular density indices on a monitor.

10. A data processing and safety control method for the anti-perforation gastrointestinal endoscope device according to any one of claims 1-9, characterized in that, This includes the following steps performed by the processor: It receives in real time the circumferential stress electrical signal of the mirror body, multispectral image pixel data, and mirror body displacement vector transmitted by the sensing integration module; Texture abrupt change features in the image are extracted by frequency domain transformation, and the relative displacement field of the tissue is calculated by optical flow method to quantify mechanical strain rate. The stress electrical signal and strain rate are weighted using the Transformer model to generate a dynamic risk score that reflects the degree of mechanical overload of the mirror body. Based on the rating level, an electrical signal is automatically sent to the actuator to trigger corresponding tactile feedback, propulsion speed limitation, or mirror body physical stiffness switching actions; when the real-time risk rating is critical risk, the electromagnetic actuator is triggered to open the emergency brake latch, physically restricting the mirror entry operation.

Images