Intelligent suturing system and method for cesarean scar uterus

Abstract

The invention discloses an intelligent suturing system and method for a caesarean scar uterus, and aims to solve the technical problems of low suturing precision and high complication risk caused by scar tissues. The system comprises a data acquisition module for acquiring mechanical and image data through an intelligent suture needle integrated with a sensor and an image acquisition device; the processing module is used for segmenting a scar boundary by adopting a deep learning model and fusing data to generate an individualized stitching strategy; the mechanical execution module is used for executing stitching according to the strategy; and the augmented reality navigation module is used for superposing virtual information on the reality view for guiding. And the processing module is also used for performing closed-loop feedback adjustment on the suture operation based on the mechanical data monitored in real time. Through closed-loop control of multi-mode perception and AI decision, individualized precise suturing of scar tissue is achieved, and the operation safety and effect are remarkably improved.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of medical devices and artificial intelligence, and in particular to a suture enhancement system and method for intelligent suturing of cesarean section scar uterus that integrates deep learning semantic segmentation technology, multimodal sensor data fusion technology, dynamic speed regulation algorithm, augmented reality navigation technology, and robot closed-loop feedback control technology. Background Technology

[0002] In recent years, with the development of medical intelligence, various advanced technologies have been applied to complex surgical scenarios. For example, deep learning semantic segmentation algorithms (such as DeepLabV3+) are used to process medical images to achieve automatic identification of specific tissues; in the field of surgical instruments, intelligent staplers and surgical robots that integrate basic force feedback or adopt TRS dynamic speed regulation algorithms have emerged to improve the accuracy of operations; in addition, augmented reality (AR) technology has also been introduced into surgical navigation, providing visual guidance to doctors by overlaying preoperative images on the intraoperative field of view.

[0003] However, in existing technologies, the application of these advanced technologies is often fragmented and isolated, failing to form a complete end-to-end solution that operates collaboratively from perception and decision-making to execution. This fragmentation of technology application means that traditional and existing intelligent assisted suturing solutions still face inherent limitations when dealing with the complex scarred uterine environment of multiple cesarean sections.

[0004] The shortcomings of existing technologies are mainly reflected in the following aspects: First, due to the lack of a unified decision-making model that can deeply integrate and correlate multi-source and multi-modal information such as real-time imaging and real-time mechanical sensing data, the phenomenon of "information silos" remains serious. Doctors' decision analysis still requires subjective integration among different information sources, lacking a "smart brain" that can provide a global, real-time, and optimal suturing strategy. Second, due to the lack of close closed-loop linkage between the perception system and the execution system, the system's intervention capabilities are mostly limited to providing auxiliary information or executing preset parameters. They cannot make truly proactive and dynamic adjustments to suturing parameters (such as depth and speed) based on real-time and accurate perception of tissue tension, which makes the suturing quality still highly dependent on the doctor's on-site feel and experience. In addition, when faced with sudden intraoperative situations such as sudden changes in tissue tension or abnormal tissue clamping by the suture device, the system also lacks a unified and rapid risk verification and emergency response mechanism, resulting in a delayed and inefficient response to abnormal events, missing the best intervention opportunity, and failing to effectively avoid serious complications such as tissue tearing.

[0005] Therefore, there is an urgent clinical need for a new generation of suture enhancement architecture that can deeply integrate multiple advanced technologies to form a synergistic effect of "perception-decision-execution" in order to fundamentally solve the problems of precision, safety and individualization in suturing of scarred uterus in multiple cesarean sections. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide an intelligent suturing system and method for cesarean section scarred uterus, aiming to at least partially solve the problems of low surgical suturing accuracy, poor safety and lack of individualization caused by information silos, lack of closed-loop linkage between perception and execution and lack of emergency response mechanism in the prior art.

[0007] This invention provides an intelligent suturing system for cesarean section scar uterus, comprising: The data acquisition module is used to acquire imaging data of the surgical area and to acquire multimodal mechanical sensing data through a smart suture needle that integrates pressure and / or strain sensors. The processing module is used to segment the boundary of scar tissue in real time based on the image data using a deep learning model, and to fuse the boundary information of the scar tissue with the mechanical sensing data to dynamically generate an individualized suturing strategy that includes parameters such as suture trajectory, suture depth and / or suture firing speed. The mechanical execution module is used to drive the robotic arm and the stapler to perform precise suturing operations according to the individualized suturing strategy. An augmented reality navigation module is used to overlay the boundary of the scar tissue and the suture trajectory as virtual information onto the real-time view of the surgical area for the operator to observe; wherein, the processing module is further configured to continuously monitor the mechanical sensing data during the suturing operation, and automatically and in real-time adjust the suturing operation when one or more preset risk conditions are met.

[0008] Optionally, during runtime, the deep learning model has an encoder and a decoder that perform the following operations: the encoder processes the input features and extracts multi-scale features through a multi-scale dilatation attention mechanism and a sliding window attention mechanism; and the decoder combines global and local features through a multi-feature channel attention fusion mechanism to generate the boundary of the scar tissue.

[0009] Optionally, when generating the individualized suturing strategy, the processing module is further configured to: estimate the elastic modulus Es of the scar tissue based on preoperative elastography data or intraoperative stress-strain curves; and determine the suturing firing speed according to a preset piecewise function relationship related to the elastic modulus Es.

[0010] Optionally, the processing module is configured to determine whether one or more preset risk conditions are met; wherein the one or more preset risk conditions include at least: the pressure value derived from the mechanical sensing data continuously exceeds a preset first pressure threshold within a preset time period, and its strain value exceeds a preset second strain threshold.

[0011] Optionally, the processing module is further configured to plan an initial suture path that avoids densely vascularized areas based on preoperative medical imaging data using the Dijkstra algorithm or a variant thereof, and to perform B-spline interpolation smoothing on the initial suture path.

[0012] This invention also provides a corresponding suture enhancement method, including steps such as acquiring images and mechanical data, segmenting scar boundaries using a deep learning model, fusing data to generate a suture strategy, guiding the suture process using an augmented reality device, controlling a mechanical execution module to perform the suture, and performing closed-loop feedback adjustments based on mechanical data during execution.

[0013] Optionally, in the method, the step of processing the image data using a deep learning model specifically includes: extracting multi-scale features of the image data using an encoder within the deep learning model and employing a multi-scale dilatation attention mechanism and a sliding window attention mechanism; and generating the boundary of the scar tissue by combining global features and local features using a decoder within the deep learning model and employing a multi-feature channel attention fusion mechanism.

[0014] Optionally, in the method, the step of generating the individualized suturing strategy further includes: estimating the elastic modulus Es of the scar tissue; and determining the suturing firing rate based on a preset piecewise function relationship associated with the elastic modulus Es.

[0015] Optionally, in the method, the preset risk conditions include at least: the pressure value derived from the mechanical sensing data is continuously greater than a first pressure threshold within a preset time period, and the strain value in the mechanical sensing data is simultaneously greater than a second strain threshold.

[0016] Optionally, before generating the individualized suturing strategy, the method further includes: planning an initial suturing path that avoids densely vascularized areas using the Dijkstra algorithm based on preoperative medical imaging data; and performing B-spline interpolation smoothing on the initial suturing path.

[0017] The suture enhancement system and method provided in this invention effectively solve the problems of isolated surgical information, subjective decision-making, and delayed risk intervention in the prior art, and have significant beneficial effects.

[0018] Specifically, this system achieves end-to-end intelligent intervention for suturing complex scarred uteruses through a collaborative architecture of "perception-decision-execution".

[0019] First, by integrating a multimodal sensor onto the suture needle and deeply fusing its real-time data with image information segmented by a deep learning model, the information silos between "vision" and "touch" are fundamentally broken down, providing a high-quality data foundation for the generation of global and precise suture strategies.

[0020] Secondly, this system utilizes a dynamic speed regulation algorithm based on individualized biomechanical parameters of patients (such as elastic modulus) and uses real-time mechanical sensing data as input for closed-loop feedback to construct a decision-making closed loop that combines foresight and real-time performance. This transforms suturing decisions from a passive, doctor-experience-based "art" into an active, real-time data-driven "science".

[0021] Furthermore, by introducing a parallel dual-verification risk monitoring mechanism, it enables second-level detection and immediate response to risks such as potential tissue tearing, transforming the traditional model that relies on doctors for post-event remediation into a new model of proactive prevention and immediate intervention.

[0022] Through the synergistic effect of the above-mentioned key technical links, the present invention system overcomes the limitations of traditional suturing methods, realizes comprehensive perception, precise planning, adaptive adjustment and intelligent risk control of the suturing process, and significantly improves the safety and effectiveness of high-risk gynecological surgeries such as multiple cesarean sections. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of this application 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 embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 A schematic block diagram illustrating the main functional modules of a suture reinforcement system provided in an embodiment of this application is shown.

[0025] Figure 2 A flowchart illustrating a suture reinforcement method provided in an embodiment of this application is shown schematically.

[0026] Figure 3 A schematic diagram illustrating a layered system architecture of a suture reinforcement system provided in an embodiment of this application is shown. Detailed Implementation

[0027] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0028] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0029] It should be noted that various aspects of the embodiments described below are within the scope of the appended claims. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the apparatus and / or practice the method. Additionally, this apparatus and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.

[0030] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The drawings only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0031] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0032] This invention provides an intelligent suturing system and method for uterine suture after cesarean section scarring. The system and method provided in this embodiment can be executed by one or more computing devices, which can be implemented as software or a combination of software and hardware, for example, integrated into a dedicated medical device, server, or terminal device.

[0033] Please see Figure 1The diagram schematically illustrates the main functional modules of the suture reinforcement system provided by an embodiment of this application. In a preferred embodiment, the system may include the following modules: Data acquisition module 100 This module is the comprehensive perception cornerstone of the entire intelligent enhancement system. Its core function is to execute a comprehensive data acquisition process that spans both preoperative preparation and intraoperative procedures. This module aims to comprehensively acquire the precise state of the surgical environment from multiple key dimensions, including "vision," "tactile," and "historical records," providing high-quality, multimodal data input for subsequent processing modules 200 to perform AI model segmentation, fusion analysis, and individualized suturing strategy generation.

[0034] In one specific embodiment, the operation of the data acquisition module 100 can be divided into preoperative data acquisition and intraoperative real-time data acquisition: Preoperative data acquisition: Before the surgery begins, the system collects individualized static data from the patient to construct a personalized biomechanical model. This data includes: Preoperative medical imaging and biomechanical data: This data is preferably acquired using high-resolution medical imaging equipment, particularly magnetic resonance imaging (MRI). Specifically, T2-weighted sequences and three-dimensional volumetric scanning techniques can be used to achieve multi-contrast imaging of the uterine scar area, thereby accurately delineating the scar's boundaries, thickness, and spatial relationship with surrounding tissues. Crucially, during MRI scanning, elastography sequences can be used to acquire elastic modulus distribution data of the scar area. This elastic modulus data quantifies tissue stiffness and is a key prior parameter for the subsequent processing module 200 to generate individualized dynamic speed adjustment strategies.

[0035] Patient's past surgical history data: The system is also configured to retrieve relevant historical information from the patient's electronic medical record (EMR). This information may include, but is not limited to: the number of previous cesarean sections, the surgical procedure used in the previous surgery (e.g., transverse or longitudinal incision in the lower uterine segment), the healing grade of the postoperative incision, and whether complications such as scar diverticulum or pelvic adhesions have occurred. By establishing a mapping relationship between these individual characteristics and scar biomechanical parameters (such as elastic modulus and viscosity coefficient), the system can build a more accurate initial model for each patient.

[0036] Intraoperative real-time data acquisition: During the surgery, the system performs dynamic, real-time multimodal data acquisition. This data includes: Real-time image data: This function is achieved by an image acquisition device deployed within the surgical area, such as a 4K laparoscope. It is responsible for continuously acquiring high-definition video or image sequences containing the uterine scar area in real time throughout the entire surgical procedure and transmitting this image data uninterruptedly to the processing module 200.

[0037] Real-time biomechanical sensing data: This function is achieved by a specially designed smart suture needle. To digitize the invisible tissue stress conditions, multiple micro-sensors are integrated onto the surface of the smart suture needle, forming a distributed sensing network along the needle's axial and circumferential directions. When the suture needle interacts with uterine tissue through contact, puncture, or lifting, this sensing network can collect data reflecting the tissue's biomechanical properties in real time. Specifically, this network may include: Pressure sensors: For example, multiple miniature piezoresistive pressure sensors (such as the FSR402 model) can be integrated on the surface of the needle body to measure the local contact stress at various points of the needle body when it comes into contact with tissue in real time.

[0038] Strain sensors: For example, a fiber Bragg grating (FBG) strain gauge can be fixed in the centerline of the needle body to monitor the overall bending or tensile deformation of the needle body under stress in real time.

[0039] In this way, the data acquisition module 100 can combine the visual and tactile information that can only be relied upon by doctors' subjective perception in traditional surgery with the objective archival data obtained before surgery, and transform them into a highly rich multimodal data stream that can be accurately analyzed by computers.

[0040] Processing module 200 Please continue reading. Figure 1 The processing module 200 is the core and "intelligent brain" of the suture reinforcement system of this invention. Physically, it is typically housed by one or more high-performance edge computing devices (e.g., embedded systems equipped with NVIDIA Jetson AGXOrin series chips) to meet the stringent requirements of low-latency, high-throughput data processing during surgery. The processing module 200 is coupled to the data acquisition module 100, responsible for receiving the multimodal data streams it acquires, and completing the entire process from data analysis and intelligent decision-making to closed-loop control through a series of complex algorithms and logic.

[0041] In one specific embodiment, the processing module 200 is configured to perform the following core functions: Scar tissue boundary segmentation based on a deep learning model: This function is the visual foundation for achieving precise suturing. The processing module 200 receives real-time image data transmitted by the data acquisition module 100 and calls a pre-trained deep learning model to process it, so as to accurately identify and segment the boundaries of scar tissue of interest to the doctor in the image in real time.

[0042] To achieve high-precision segmentation, this deep learning model preferably employs an encoder-decoder structure. Specifically, the encoder's core task is to extract rich features from the input image. To efficiently capture the visual features of scar tissue at different scales (e.g., from tiny scar points to large fibrotic areas), a multi-scale dilation attention mechanism is preferred. This mechanism can be implemented by setting multiple parallel processing branches with different dilation rates. Simultaneously, to reduce computational complexity while maintaining a large receptive field to meet real-time requirements, the encoder can also employ a sliding window attention mechanism, restricting global attention computation to a local window. The decoder's task is to restore the abstract features extracted by the encoder into pixel-level, clear segmentation results. To make the output scar boundaries more refined and accurate, the decoder preferably employs a multi-feature channel attention fusion mechanism. This mechanism can intelligently combine feature information from different levels. For example, it can weightedly fuse high-level global features containing location and context information with low-level local features containing texture and gradient details, thereby enhancing the response to pixels in the boundary region and ultimately generating a high-precision scar tissue boundary with an error of less than 0.5 mm.

[0043] Data Fusion and Personalized Suturing Strategy Generation: After obtaining a clear scar boundary, the core task of processing module 200 is to make intelligent decisions and generate a suturing strategy "tailor-made" for the current patient. Data Fusion: The data fusion module plays a central, pivotal role within processing module 200. As a crucial bridge connecting "perception" and "decision-making," its main task is to receive and fuse heterogeneous data streams from different sources to generate a unified, accurate, and high-dimensional real-time status understanding, providing a high-quality data foundation for the subsequent generation of personalized suturing strategies.

[0044] In one specific embodiment, the data fusion module performs the following key operations: Multimodal Data Fusion: The core function of this module is to fuse two completely different types of data: Geometric anatomical information: namely, the boundary information of scar tissue segmented in real time by a deep learning model. This data tells the system "where is the scar." Biomechanical information: namely, the mechanical sensing data collected in real time from the sensor network of the smart suture needle. This data tells the system "how much force" and "how much deformation." Spatio-temporal Registration: Because the sampling frequencies of image data (e.g., 30Hz) and sensor data (e.g., 100Hz) are different, and the coordinate systems also need to be unified, simple superposition is ineffective. Therefore, the data fusion module must perform spatio-temporal registration. This means that it can use complex algorithms to accurately and one-to-one correlate the mechanical readings at a specific moment (time dimension) with the exact location of the suture needle at that moment (spatial dimension) and the type of tissue it contacts (anatomical dimension, i.e., whether it is a scar). Optimal state estimation (using the extended Kalman filter algorithm): In order to obtain the most stable and accurate fusion results in a real surgical environment full of noise and uncertainty, the data fusion module preferably uses the extended Kalman filter (EKF) algorithm for optimal state estimation.

[0045] EKF is a powerful recursive algorithm that excels at estimating the state of a dynamic system from a series of incomplete and noisy measurements. In this invention, the "dynamic system" is the interaction state between the suture needle and the tissue. The state vector of EKF can be defined as including the three-dimensional coordinates and orientation of the suture needle tip, as well as the three-dimensional tension components it experiences. The observation equation includes the boundary distance obtained from image segmentation and the pressure / strain values ​​obtained from the sensor. Through continuous "prediction-update" loops, EKF can filter out noise, correct errors, and finally output a fused, high-confidence state estimate with a position estimation error of less than 0.3 mm and a tension measurement error of less than 0.05 N.

[0046] In this way, the data fusion module "extracts" two independent, noisy raw data streams into a unified, clean, and highly reliable real-time status report, which greatly improves the accuracy and reliability of subsequent stitching strategy generation.

[0047] Individualized strategy generation: Based on the fused data and combined with the preoperative data obtained from the data acquisition module 100, the processing module 200 can dynamically generate a set of individualized suturing strategies that include multiple dimensions.

[0048] Dynamic speed control: Based on the elastic modulus Es of the scar tissue obtained preoperatively, the system determines the firing speed of the suture device according to a preset piecewise function relationship related to Es (e.g., a specific TRS dynamic speed control algorithm). The general trend is that the suture firing speed decreases inversely as the elastic modulus Es increases. Path planning: The system can also plan a safe initial suture path based on vascular distribution information in preoperative MRI images, using the Dijkstra algorithm or its variants, while avoiding densely vascularized areas. To ensure smooth and stable execution by the robotic arm, the system also performs B-spline interpolation smoothing on this path. In addition, the strategy includes optimal suture depth, stitch length, and other parameters.

[0049] Risk condition assessment and closed-loop feedback control: This is the core innovation of this invention that distinguishes it from traditional auxiliary systems.

[0050] Risk Condition Judgment: During the suturing process, the processing module 200 continuously monitors real-time mechanical sensing data and determines whether it meets one or more preset risk conditions. A preferred judgment method that effectively avoids false alarms employs a dual verification condition: Condition 1 (State Judgment): The pressure value derived from the mechanical sensing data continuously exceeds a preset first pressure threshold (e.g., 1.5N) for a short preset time period (e.g., 30 milliseconds). This "continuously" limitation is to filter out instantaneous signal noise and ensure that a stable and real overload state is detected.

[0051] Condition 2 (Event Judgment): Simultaneously with satisfying Condition 1, the strain value derived from the mechanical sensing data also instantaneously exceeds a preset second strain threshold (e.g., 0.02%). Exceeding the strain limit usually indicates that the needle structure has undergone dangerous deformation, and is a high-weight risk indicator.

[0052] Closed-loop feedback control: Only when the above dual verification conditions are met simultaneously will the processing module 200 finally confirm the occurrence of the risk and immediately generate an adjustment command to send to the mechanical execution module 300 to automatically and in real time adjust the current suturing operation. For example, the PID closed-loop control algorithm can be used to precisely adjust the suturing depth back by 0.5mm, thereby proactively and immediately avoiding the risk of tissue tearing.

[0053] Finally, the processing module 200 will convert the generated complete suturing strategy (trajectory, speed, depth, etc.) into low-level execution control instructions and send them to the mechanical execution module 300; at the same time, it will send the virtual information (scar boundary, suturing trajectory) used for navigation to the augmented reality navigation module 400.

[0054] Mechanical Actuation Module 300 Please continue reading. Figure 1 The mechanical execution module 300 is the "precision arm" of the system of this invention, and is the physical terminal that realizes all intelligent decisions and suturing strategies. It is tightly coupled to the processing module 200 and is responsible for receiving and faithfully executing the individualized suturing strategies and low-level control instructions issued by it.

[0055] In one specific embodiment, the mechanical actuation module 300 mainly consists of a multi-degree-of-freedom robotic arm and a smart stitcher mounted at its end. Its design goal is to achieve sub-millimeter-level operational accuracy and high operational stability.

[0056] The robotic arm's precise positioning function: The robotic arm is responsible for precisely transporting the intelligent suture needle and suture device to the target position and posture planned by the processing module 200.

[0057] To achieve high-precision positioning, the joints of the robotic arm are preferably driven by high-performance servo motors. These servo motors can be equipped with high-resolution absolute encoders (e.g., 16-bit or higher) to ensure that the system can accurately know the angle of each joint in real time, thereby achieving a repeatability of ±0.3mm or even higher.

[0058] Furthermore, to avoid accidental rigid collisions and damage to tissues during movement, a six-dimensional force / torque sensor can be integrated into the end effector or wrist of the robotic arm. This allows the processing module 200 to monitor the forces acting on the robotic arm and trigger a passive compliant control strategy when necessary, enabling the robotic arm to exhibit flexibility when encountering unexpected forces, thereby ensuring surgical safety.

[0059] The intelligent suture device has precise operation functions: The intelligent suture device is responsible for performing the final suturing action. Its core function is to precisely control the suturing process based on the dynamic parameters issued by the processing module 200.

[0060] Suture depth control: To precisely control the tissue penetration depth of each suture, the depth control mechanism of the suture device preferably employs a combination of a stepper motor and a ball screw mechanism. This combination can convert the rotational motion of the motor into extremely fine linear motion, achieving micron-level adjustment of the suture depth.

[0061] Control of suture firing speed: To execute the dynamic firing speed calculated by the processing module 200 based on the elastic modulus Es of scar tissue, the suture's drive system can integrate an electromagnetic clutch or employ a variable-speed motor. This allows the system to switch the transmission ratio or adjust the motor speed in real time according to instructions, thereby achieving dynamic adjustment of the firing speed.

[0062] Abnormal Tissue Detection: The jaws of the intelligent suture device can also integrate a pressure sensor to determine whether abnormally thick tissue or foreign objects are clamped during closure. Once an abnormality is detected, the suture device will refuse to execute the firing command and will report the abnormal status to the processing module 200.

[0063] Closed-loop feedback and safety braking: The mechanical actuation module 300 is not only a passive command executor, but it also participates in the closed-loop feedback process.

[0064] Operational feedback: Throughout the operation, the encoder, six-dimensional force sensor, and various sensors on the sewing machine (such as laser displacement sensor) of the robotic arm will transmit their position, force, status and other information back to the processing module 200 in real time, providing data for the upper-level PID error correction and global status monitoring.

[0065] Safety braking: Once the mechanical actuation module 300 receives the highest priority abort command from the processing module 200 (e.g., when a serious risk is detected), the drive module will cut off the power to all motors and lock the robotic arm joints within a very short time (e.g., 0.1 seconds) to maximize the safety of the patient and the equipment.

[0066] Through the above-mentioned meticulous design, the mechanical execution module 300 ensures that the complex and dynamic stitching strategy generated by the processing module 200 can be implemented in physical reality without compromise and safely.

[0067] Augmented Reality Navigation Module 400 Please continue reading. Figure 1 The augmented reality navigation module 400 serves as an interactive bridge and visual terminal connecting "machine intelligence" and "human doctors" in this invention system. It is tightly coupled to the processing module 200, and its core function is to transform the abstract data and decision instructions generated by the processing module 200 into visual guidance information that the operator (doctor) can intuitively understand and utilize, thereby seamlessly combining the computational advantages of artificial intelligence with the clinical experience and decision-making ability of human doctors.

[0068] In one specific embodiment, the augmented reality navigation module 400 mainly consists of an AR device (e.g., a head-mounted display such as the Microsoft HoloLens 2) and navigation software running on it. It receives virtual information sent by the processing module 200 and performs the following key operations: Receiving and visualizing virtual information: The virtual information received by this module mainly includes: coordinate data of the scar tissue boundary segmented by a deep learning model. The data received by this module mainly includes: data representing the scar tissue boundary: This is typically a sequence of three-dimensional spatial coordinate points segmented by a deep learning model that defines the boundary contour. Data representing the suture trajectory: This is also a sequence of three-dimensional spatial coordinate points generated by a path planning algorithm that defines the recommended suture path. After receiving these two sets of coordinate data sequences, the navigation software performs rendering operations. For example, by interpolating between the "boundary" coordinate points, a continuous, high-brightness virtual boundary line is generated; and by interpolating between the "trajectory" coordinate points, a series of virtual arrows or path indicator lines representing the needle insertion point and needle exit direction are generated. Then, these rendered virtual graphics are precisely and in real time overlaid onto the actual surgical area in the operator's field of vision.

[0069] The optimal suture trajectory, planned by the processing module 200, includes a series of anchor points and smooth curves connecting them. Real-time mechanical state markers, such as tension over-limit areas determined by risk condition judgment logic, are also included. Upon receiving this data, navigation software (e.g., developed using the Unity engine) renders it in an intuitive graphical format. For example, scar tissue boundaries are presented as a red semi-transparent mask, the planned suture trajectory as a yellow highlighted curve, and tension over-limit areas as a flashing yellow warning. High-precision virtual-real fusion and spatial registration: To ensure perfect alignment between virtual information and the real surgical scene, this module executes a complex spatial registration algorithm.

[0070] Initial registration and distortion correction: At the start of the surgery, the system performs distortion correction on the real-time transmitted images based on pre-calibrated laparoscopic lens parameters to ensure image fidelity. Subsequently, by identifying markers on surgical instruments or using other calibration methods, the virtual information coordinate system is precisely spatially registered with the laparoscopic image coordinate system, achieving pixel-level alignment.

[0071] Depth Perception Simulation: To enable doctors to intuitively judge the angle of needle insertion and the depth of tissue penetration, this module can use a ray casting algorithm to simulate the depth perception effect of the suture needle trajectory, making the two-dimensional virtual guide line present a three-dimensional spatial sense.

[0072] Dynamic registration correction: Considering that tissues may shift during surgery due to breathing, instrument manipulation, etc., leading to initial registration failure, this module integrates an inertial measurement unit (IMU) and visual odometry to track the position and orientation of surgical instruments and endoscopes in real time. Once significant displacement is detected, a dynamic registration correction algorithm is triggered to maintain spatial consistency between virtual markers and physical anatomical structures.

[0073] Human-computer collaborative interaction: This module provides doctors with flexible interaction capabilities to ensure that doctors are always in a dominant position.

[0074] View adjustment: Doctors can adjust the transparency of the AR view in real time through gesture interaction or voice commands, or switch between different display modes (e.g., displaying the scar boundary mode alone, overlaying the suture path mode, or displaying the real-time tension heat map mode).

[0075] Strategy fine-tuning: Doctors can also fine-tune the suturing strategy generated by the processing module 200 under AR guidance. For example, by using gestures or force feedback handles, they can drag and adjust the position of a suturing anchor point in the AR view. After receiving this fine-tuning instruction, the processing module 200 will automatically recalculate and smooth the suturing trajectory.

[0076] Visualized risk warning communication: When the processing module 200 determines that the risk conditions are met, the augmented reality navigation module 400 will immediately convey the risk to the doctor in the most prominent way. For example, it will automatically highlight the virtual boundary of the risk area in red and make it flash, while at the edge or corner of the field of vision, it will display key prompts such as the current tension value and depth error in real time in the form of numbers or curves, to help the doctor make the most accurate intervention decision in a very short time.

[0077] Through a mechanism of "real-time data-driven, precise virtual-real integration, and flexible interactive response," the Augmented Reality Navigation Module 400 transforms abstract mechanical parameters and complex AI strategies into intuitive, operable, and interactive spatial guidance information, thereby significantly improving the accuracy, safety, and efficiency of human-machine collaboration in surgical procedures.

[0078] This embodiment provides a method for strengthening the intelligent suturing of a cesarean section scarred uterus. This method can be executed by the system described in Embodiment 1 to achieve precise and safe suturing of a complex scarred uterine environment.

[0079] Optionally, to achieve high-precision scar tissue boundary segmentation, the deep learning model deployed in the processing module 200 preferably employs an encoder-decoder structure. This structure has proven to be efficient and mature in the field of image segmentation. In a specific embodiment of the present invention, the internal workings of the model are as follows: Regarding the encoder and its operation: The core task of the encoder is to extract rich, multi-level feature information from the input image. In order to efficiently capture the visual features of scar tissue at different scales (e.g., from tiny scar points to large fibrotic areas) without sacrificing the receptive field (i.e., the model's field of view), it preferably employs a multi-scale dilation attention mechanism. Specifically, the input image feature map can be fed into multiple parallel processing branches, each branch employing a different dilation rate, such as 1, 2, 3, to capture different levels of features from fine textures to large-scale morphologies. Simultaneously, to address the problem of excessive computational cost in processing high-resolution images using traditional Transformer models and to meet the real-time requirements of the surgical procedure, the encoder can also employ a sliding window attention mechanism. This mechanism confines global attention computation to a series of overlapping local windows, thus significantly reducing computational complexity while maintaining effective modeling of contextual information. The encoder ultimately outputs a set of multi-scale features containing rich contextual information. Regarding the decoder and its operation: the core task of the decoder is to restore the abstract features extracted by the encoder into pixel-level segmentation results with clear boundaries. To make the output scar boundaries more refined and accurate, the decoder preferably employs a multi-feature channel attention fusion mechanism. This mechanism intelligently combines feature information from different sources for the final pixel-level judgment. Specifically, it combines: multi-scale features generated by the encoder; global features obtained through analysis of the entire image, such as the approximate location and proportion of the scar in the entire field of view; and local features calculated for each pixel, such as gradients, textures, and other detailed information at that point.

[0080] By dynamically and selectively weighting and fusing these features, the decoder can simultaneously consider both the "big picture" and the "details," thereby enhancing the response to pixels in the real boundary region and ultimately generating high-precision scar tissue boundaries.

[0081] Optionally, to achieve a more in-depth, patient-individualized suturing strategy, the processing module 200 is also configured to perform a key dynamic speed adjustment function when generating the individualized suturing strategy. The core of this function is to no longer use a fixed suture firing speed, but instead to adaptively adjust the speed in real time based on the biomechanical characteristics of the tissue at the suture point.

[0082] In a specific embodiment, the dynamic speed regulation function is implemented as follows: Step 1: Estimate the elastic modulus Es of the scar tissue. This parameter is a key indicator for quantifying tissue "stiffness". The processing module 200 can acquire or estimate it in two ways: Based on preoperative data: Using preoperative elastography MRI data acquired from the data acquisition module 100, a priori map of the elastic modulus distribution in the patient's scar area can be directly obtained. Based on intraoperative data: During suturing, by analyzing the stress-strain curve fed back in real time by the sensors on the smart suture needle, the Stokes modulus or Young's modulus of the currently contacting tissue can be dynamically calculated, thereby updating the Es value in real time. For example, the slope of the curve can be calculated using the formula k = Δσ / Δε, where: k is the calculated instantaneous stiffness; Δσ is the change in stress measured by the pressure sensor within a very small time period. Stress reflects the force per unit area within the tissue; Δε is the change in strain measured by the strain sensor within the same time period. Strain reflects the degree of relative deformation of the tissue under stress. An update strategy of Es = 0.8 * Es_old + 0.2 * k is used, where Es is the updated elastic modulus estimate, and Es_old is the original elastic modulus estimate. This achieves adaptive adjustment. The second step involves determining the firing speed based on the elastic modulus Es. After obtaining the elastic modulus Es, the processing module 200 determines the optimal firing speed of the suture needle based on a preset piecewise function relationship related to Es. This function relationship can be a specific implementation of a TRS (Tissue-adaptive Reload System) dynamic speed adjustment algorithm. This function relationship follows a core principle: the suture firing speed decreases inversely as the elastic modulus Es increases. This is because a slower, more stable firing speed is required on harder, more fibrous scar tissue to ensure that the suture needle has sufficient time to penetrate the tissue and achieve perfect shaping, thereby avoiding poor nail formation or tissue tearing due to excessive speed. For example, this piecewise function relationship can be set as follows: when Es is less than a first hardness threshold (e.g., 200 kPa, representing softer tissue), a higher conventional speed (e.g., 20 mm / s) is used. When Es is between the first and second hardness thresholds (e.g., 200-500 kPa), the speed is continuously adjusted according to a linear or non-linear function, for example, v = 12 + 0.04 * (Es - 200) mm / s. When Es is greater than a second hardness threshold (e.g., 500 kPa, representing very hard tissue), a lower safety speed is used. In this way, the system upgrades the suturing operation from a "one-size-fits-all" mode to a "personalized" intelligent operation mode that can precisely sense and adapt to the tissue characteristics of each suture point, thereby significantly improving suturing quality and safety.

[0083] Optionally, to achieve robust risk assessment that effectively avoids false alarms, the processing module 200, when executing closed-loop feedback control logic, preferably includes a specially designed dual-verification condition among the "one or more preset risk conditions" it assesses. This condition is not based on a single sensor reading, but is triggered only when two sub-conditions of different natures are met simultaneously, thereby greatly improving the accuracy and reliability of risk assessment.

[0084] In a specific embodiment, the logic of the dual verification condition is as follows: The first sub-condition (state judgment): This condition determines whether a stable, continuous overload state exists. Specifically, it requires that the pressure value from the mechanical sensing data continuously exceeds a preset first pressure threshold (e.g., 1.5N) for a short, predetermined time period (e.g., 30 milliseconds). It is important to note that the introduction of the "continuous" time dimension is intended to intelligently filter out transient signal noise or "glitch" caused by current fluctuations or minor, rapid scraping of the instrument, ensuring that the system only responds to stable, high-pressure states representing actual tissue overload. The second sub-condition (event judgment): This condition determines whether a critical, high-risk physical event has occurred. Specifically, it requires that, while satisfying the first sub-condition, the strain value from the mechanical sensing data also instantaneously exceeds a preset second strain threshold (e.g., 0.02%). Exceeding the strain limit typically means that the intelligent suture needle's needle structure itself has undergone significant and dangerous bending or tensile deformation due to excessive force. This is itself a high-weight risk indicator, and its emergence greatly enhances the certainty of risk assessment.

[0085] Only when both of the above sub-conditions—namely, a stable pressure overload state and a critical structural deformation event—are simultaneously met will the processing module 200 finally confirm the occurrence of the risk and trigger subsequent automatic adjustment operations, such as callback stitching depth. This asymmetric dual-verification logic elevates the system's risk perception capability from simple threshold alarms to a new level of intelligent judgment capable of distinguishing between "interference" and "real danger."

[0086] Optionally, to achieve safer and more proactive suture path planning, the processing module 200 is also configured to perform a path planning step based on preoperative data before generating the final individualized suture strategy. This step aims to leverage the advantages of computer vision and path search algorithms to plan an optimal suture trajectory that proactively avoids known risk areas and facilitates smooth robotic execution before the surgery begins.

[0087] In a specific embodiment, the path planning step is implemented as follows: The first step is to plan the initial suture path based on preoperative data, preferably based on preoperative medical imaging data, such as MRI angiography (MRA) markers acquired in the data acquisition module 100. This is because preoperative images are usually clearer and have fewer artifacts, and can more reliably identify important, densely vascularized areas that need to be avoided, such as the uterine artery, than intraoperative real-time images. At the algorithm level, the system can model the uterine tissue area to be sutured as a graph in three-dimensional space, where densely vascularized areas are marked as high-cost or impassable regions. Subsequently, the system uses the classic Dijkstra algorithm or its variants (such as the A* algorithm) to search and calculate the lowest-cost (i.e., safest and optimal) path in the graph, with the suture start and end points as targets. This algorithm-generated path that perfectly avoids known "danger zones" is the initial suture path. The second step: Smoothing the initial suture path. It's important to note that the initial suture path directly generated by the Dijkstra graph search algorithm is mathematically a series of discrete straight line segments forming a broken line, with sharp inflection points at the connection points. Directly issuing such a path instruction to the mechanical execution module 300 would cause frequent starts and stops of the robotic arm, resulting in jerky movements, low efficiency, and the mechanical impact during acceleration and deceleration could cause additional tensile damage to fragile tissues. To address this issue, the system needs to smooth this initial suture path using B-spline interpolation. B-spline interpolation is an advanced mathematical tool that can fit this series of discrete broken line points into a smooth curve that is mathematically continuous and has a gentle curvature. This smoothed curve is the final suture trajectory issued to the mechanical execution module 300 for execution.

[0088] By employing this two-step planning method of "first using Dijkstra's algorithm to mitigate macroscopic risks, and then using B-splines to optimize microscopic motions," this invention ensures that the suturing operation combines high safety with smooth execution.

[0089] Please see Figure 2 The diagram illustrates, schematically, the process executed by the suture reinforcement method provided in this application. In a preferred embodiment, the method may include the following steps: Step S10: Data Acquisition. This step is the starting point of the entire method. After the system starts, corresponding to... Figure 1The data acquisition module 100 begins operation, performing data acquisition operations in parallel and continuously. This includes acquiring real-time image data of the surgical area through an image acquisition device, and simultaneously acquiring multimodal biomechanical sensing data in real time as the suture interacts with the uterine tissue using sensors integrated on the smart suture needle. As previously mentioned, this step may further include the acquisition and loading of preoperative data.

[0090] Step S20: Segmenting scar boundaries using a deep learning model. This step is the core of the visual processing for accurate localization. The acquired real-time image data is continuously transmitted to the processing module 200 and processed by the pre-trained deep learning model deployed therein. This model performs semantic segmentation to accurately and in real-time identify and delineate the boundaries of scar tissue from complex backgrounds. Its output (e.g., a boundary mask) will serve as an important input for subsequent steps.

[0091] Step S30: Data Fusion and Generation of Individualized Suturing Strategy. This step is the core of the method's decision-making. The processing module 200 receives the scar boundary information output from step S20 and fuses it with the real-time mechanical sensing data acquired in step S10. As mentioned above, this fusion process is spatiotemporally registered and can employ algorithms such as extended Kalman filtering to obtain optimal state estimation. Based on the fused, high-dimensional real-time state perception, and further combined with preoperative patient-specific data (such as elastic modulus Es), the processing module 200 dynamically generates a complete individualized suturing strategy. This strategy defines in detail the current optimal operating parameters, including but not limited to a safe suture trajectory, appropriate suture depth, and dynamically adjusted suturer firing speed based on tissue stiffness.

[0092] Step S40: Visual guidance via AR device. This step serves as a bridge between machine decision-making and the human doctor. The virtual information generated in Step S30 (such as calculated scar boundaries and planned suture trajectories) is sent to the augmented reality navigation module 400. This module, through its included AR device, precisely overlays this information in graphical form (e.g., semi-transparent masks, highlighted curves) onto the operator's (doctor's) real-world view, providing intuitive and clear "surgical navigation."

[0093] Step S50: Control the mechanical execution module to perform suturing. In this step, the individualized suturing strategy generated in step S30 is converted into low-level, executable control instructions by the processing module 200 and sent to the mechanical execution module 300. The robotic arm and suture device of the mechanical execution module 300 begin performing the first or next precise suturing operation according to these instructions.

[0094] Steps S60-S90: Closed-Loop Feedback and Judgment Loop. This part is the essence of the method of this invention, ensuring that each suturing operation is performed in a safe and controllable closed loop. During the suturing process (S50), the system enters a continuous monitoring and judgment loop: Step S60: Continuously monitor mechanical sensing data. During the process of the suture needle penetrating and lifting the tissue, the smart suture needle of the data acquisition module 100 continuously feeds back real-time mechanical sensing data to the processing module 200.

[0095] Step S70: Determine whether the risk conditions are met. The processing module 200 judges the real-time data fed back from step S60 according to its internally set risk judgment logic (e.g., the pressure and strain dual verification conditions detailed above).

[0096] If the judgment is "yes", meaning the preset risk conditions are met, the process will proceed to step S80.

[0097] If the result is "no", meaning the current operation is safe, the process will proceed to step S90.

[0098] Step S80: Real-time adjustment of suturing operation. In this step, the processing module 200 immediately generates and issues an adjustment command, which is then executed by the mechanical execution module 300 in real-time, automatic mode (e.g., adjusting the suture depth back by 0.5 mm). After the adjustment is complete, the process returns to normal, continues the suturing operation, and re-enters the continuous monitoring phase of step S60.

[0099] Step S90: Determine if the suturing is complete. If the operation at the current suture point is safe (i.e., step S70 is "No"), the system will further determine whether the entire suturing task has been completed.

[0100] If the result is "no", meaning there are still subsequent suture points to be processed, the process will return to step S50 to begin the operation of the next suture point.

[0101] If the result is "yes", meaning all suturing tasks have been completed, then the entire process ends.

[0102] By organically combining and iterating the above steps, the method of the present invention breaks down a complex suturing surgery into a series of controllable processes driven by data, AI decision-making, and with closed-loop feedback to ensure safety in real time.

[0103] Optionally, in step S20, the step of "processing the image data using a deep learning model" is preferably performed using a deep learning model with an encoder-decoder structure, which has been described in detail in Embodiment 1.

[0104] Specifically, this step includes: first, efficiently extracting multi-scale features from the image data using the encoder and its built-in multi-scale dilatation attention mechanism and sliding window attention mechanism; then, combining global and local features using the decoder and its built-in multi-feature channel attention fusion mechanism to finally generate a high-precision scar tissue boundary. The specific working principles of these mechanisms have been described in detail in the preceding description of processing module 200, and will not be repeated here.

[0105] Optionally, in step S30, in order to achieve a more in-depth, patient-individual-based “generation of individualized suturing strategy” step, this step also preferably includes performing a key dynamic speed adjustment function.

[0106] In one specific embodiment, this function is implemented in the following ways: First, the elastic modulus Es of the scar tissue is estimated. As described in Example 1 above, this parameter can be obtained by analyzing preoperative elastography data or by analyzing stress-strain curves fed back in real time during surgery.

[0107] Then, based on the estimated elastic modulus Es, the optimal firing speed of the suture device is determined through a preset piecewise function relationship related to Es (e.g., a specific TRS dynamic speed regulation algorithm). The core principle of this function relationship is that the suture firing speed decreases inversely as the elastic modulus Es increases, to ensure optimal suturing results in tissues of varying hardness. The specific settings of this piecewise function have been described in detail in the preceding description of processing module 200, and will not be repeated here.

[0108] Optionally, in step S70, in order to achieve robust risk assessment that can effectively avoid false alarms, the step of "determining whether the preset risk conditions are met" is preferably performed through a specially designed double verification condition.

[0109] In a specific embodiment, the judgment logic of the dual verification condition is as follows: based on the mechanical sensing data acquired in real time from the data acquisition module 100, it is determined whether two sub-conditions are simultaneously met: first, the pressure value from the mechanical sensing data continuously exceeds a preset first pressure threshold within a preset short time period; second, while meeting the first condition, the strain value from the mechanical sensing data also instantaneously exceeds a preset second strain threshold. Only when these two conditions, respectively reflecting "stable overload state" and "critical structural event," are simultaneously met, will the system ultimately confirm the occurrence of the risk and proceed to step S80 to perform subsequent automatic adjustment operations. The specific technical principles and parameter settings of this dual verification condition have been described in detail in the preceding description of the processing module 200, and will not be repeated here.

[0110] Optionally, in order to achieve safer and more forward-looking suture path planning, the method may preferably include an initial suture path planning step before step S30 "generating an individualized suture strategy".

[0111] In a specific embodiment, the path planning step includes: First, based on preoperative medical imaging data (e.g., MRI angiography markers), a safe initial suture path is planned using the Dijkstra algorithm or its variants, while avoiding known high-risk areas such as densely vascularized regions. Second, since the initial suture path is composed of a series of polygonal line segments, B-spline interpolation is performed to smooth the path and convert it into a mathematically continuous and smooth curve as the final suture trajectory, in order to facilitate smooth and stable execution by the robotic arm. The specific technical principle of this two-step planning method of "first using Dijkstra to avoid macroscopic risks, and then using B-spline to optimize microscopic motion" has been explained in detail in the previous description of the processing module 200, and will not be repeated here.

[0112] Please see Figure 3 The diagram schematically illustrates a layered system architecture of a suture reinforcement system according to another embodiment of this application. Figure 3 As shown, the entire system can be logically divided into a data layer 10, a processing layer 20, an execution layer 30, an interaction layer 40, and a security layer 50. Each layer will be described in detail below.

[0113] Data layer 10, which functionally corresponds to... Figure 1 The data acquisition module 100 is the data source for all system decisions and analyses. Figure 3 It reveals more clearly the complete composition of its data input, which together build a comprehensive cognitive foundation for the subsequent processing layer 20, containing both static prior knowledge and dynamic real-time information.

[0114] In a specific embodiment, the data sources contained in the data layer 10 can be divided into the following categories: MRI / 4K image data: As mentioned above, this part includes MRI sequences (including elastography sequences) acquired before surgery and high-definition images acquired in real time by the image acquisition device during surgery.

[0115] Intraoperative sensor signals: As mentioned above, this data comes from the distributed sensing network on the smart suture needle and reflects the mechanical properties of the tissue in real time during the interaction process.

[0116] Previous surgical history: This is in Figure 3This is a new and crucial data dimension that is particularly emphasized in the text. To achieve a deeper level of personalization, the system is configured to retrieve and integrate relevant historical information from the patient's electronic medical record (EMR). This information may include, but is not limited to: the number of previous cesarean sections, the surgical procedure used in the previous surgery (e.g., a transverse or longitudinal incision in the lower uterine segment), the healing grade of the postoperative incision, and whether complications such as scar diverticulum or pelvic adhesions have occurred.

[0117] It is important to note that the value of introducing the "past surgical history" dimension lies in its ability to establish an initial mapping relationship between individual patient characteristics and scar biomechanical parameters (such as elastic modulus and viscosity coefficient). By analyzing a large amount of case data, the system can learn prior rules such as "patients who have undergone a third cesarean section with a previous longitudinal incision are more likely to have a higher initial elastic modulus of their scar tissue." These rules help to build a more accurate and personalized initial biomechanical model for each patient, thus providing a better starting point for the generation of strategies in processing layer 20.

[0118] Processing layer 20, please continue reading Figure 3 The processing layer 20 functionally corresponds to... Figure 1 The processing module 200 is the core computing and decision-making center of the system of this invention. Figure 3 The functional division reveals the complex logic of the processing layer 20 in more detail. It uses high-performance edge computing devices as carriers to receive and process multimodal data streams from the data layer 10, and finally generate and issue control commands and navigation information.

[0119] In one specific embodiment, the internal workflow of the processing layer 20 is completed collaboratively by the following key sub-modules: The image segmentation module is a dedicated channel for processing visual information. It receives real-time image data from data layer 10 and processes it using a pre-trained deep learning model. The core task of this module is to accurately and in real-time identify and segment the boundaries of scar tissue in the image. To meet the requirements of clinical applications, the deep learning model used in this module, as described in Example 1 above, preferably employs an encoder with a multi-scale dilated attention mechanism and a decoder with a multi-feature channel attention fusion mechanism. The module ultimately outputs sub-millimeter precision boundary contour coordinate data that can be used by subsequent modules.

[0120] The signal processing module, which works in parallel with the image segmentation module, is a dedicated channel for processing mechanical information. It is responsible for receiving raw, noisy mechanical sensing data from the smart suture needle and transforming it into structured mechanical state data that can be used for decision-making.

[0121] In one specific embodiment, the module first amplifies and anti-aliasing filters the analog signal output from the sensor through a front-end conditioning circuit, and then converts it into a digital signal via a high-level ADC (e.g., a 24-bit ADC). Subsequently, a pre-trained time-series feature extraction model can be invoked to perform real-time computation on the digital signal, filter out high-frequency noise, and output key mechanical features such as tissue tension distribution heatmaps and needle deformation curvature. These mechanical features are then sent to the data fusion module for further processing.

[0122] The edge computing data fusion module, the core and central hub of processing layer 20, is responsible for fusing information from different sources and modalities and making advanced decisions based on this fusion. It receives geometric anatomical information (where the scar is) from the image segmentation module and biomechanical information (force magnitude) from the signal processing module. Through a spatiotemporal registration algorithm, this module accurately correlates these two types of information. In a preferred embodiment, the Extended Kalman Filter (EKF) algorithm can be used to perform optimal state estimation on these two types of data to obtain a unified, accurate, and high-dimensional real-time state perception. Based on the fused state, this module executes upper-level decision-making algorithms to dynamically generate individualized suturing strategies. As described in Embodiment 1 above, this includes dynamic speed regulation based on the elastic modulus Es and path planning based on Dijkstra's algorithm.

[0123] The execution control module, one of the output interfaces of processing layer 20, acts as a "translator," responsible for converting high-level decisions generated by the data fusion and strategy generation module (e.g., "stitch along trajectory A at speed V to depth D") into low-level control commands that can be directly executed by physical hardware. For example, it converts the target's three-dimensional coordinates into pulse sequences and direction signals for the servo motors of each joint of the robotic arm, and converts the stitching speed and depth parameters into specific electrical signals that drive the stepper motors or electromagnetic clutches inside the stitcher. These commands are then sent to execution layer 30.

[0124] The AR navigation module is another important output interface of the processing layer 20. It is responsible for generating virtual information for visual guidance and sending this information to the AR device in the interaction layer 40. It receives information such as scar boundaries, suture trajectory, and real-time mechanical state from the data fusion module and renders it into graphic elements (such as semi-transparent masks, highlight curves, etc.) that can be displayed by the AR device.

[0125] Through the precise division of labor and collaborative work of the above sub-modules, the processing layer 20 completes the entire intelligent computing process from raw data input to intermediate state processing, and finally to decision-making and instruction output.

[0126] Execution layer 30, please continue reading Figure 3The execution layer 30 functionally corresponds to... Figure 1 The mechanical execution module 300 in this invention is the physical operation terminal and "precision arm" of the system. Its core task is to receive and faithfully execute control commands and drive parameters from the processing layer 20, transforming the virtual, calculated suturing strategy into precise, stable, and safe physical suturing actions in the real world.

[0127] In one specific embodiment, the execution layer 30 mainly consists of a robotic arm and a stitcher. Figure 3 It clearly demonstrates the information interaction relationships between it and other layers: Receiving instructions from processing layer 20: As shown by the arrow emanating from the "execution control module" of processing layer 20, execution layer 30 receives control instructions containing suture trajectory coordinates, as well as drive parameters including firing speed, suture depth, and underlying motor drive signals. Providing operational feedback to processing layer 20: During operation, execution layer 30 is not a one-way "dumb" terminal. As shown by the "operation feedback" arrow, its internal sensors (such as robotic arm joint encoders and end effector force sensors) collect its own contact force, position, and other status information in real time and transmit this operational feedback data back to processing layer 20, providing a basis for the closed-loop control and status monitoring of the upper layer. Responding to intervention from interaction layer 40: Execution layer 30 can also respond to manual fine-tuning instructions from interaction layer 40 (doctor), making real-time corrections to its current actions, reflecting the human-machine collaborative design. Obey the highest command of safety layer 50: In an emergency, execution layer 30 will receive a stop command or alarm signal from safety layer 50 (alarm module) and immediately trigger protective measures such as emergency stop to ensure the safety and controllability of the surgical procedure.

[0128] Through this multi-directional information interaction, the execution layer 30 is not just a simple executor, but an intelligent physical operation platform that is deeply integrated into the entire perception, decision-making, and feedback closed loop.

[0129] Interaction layer 40, please continue reading Figure 3 The interaction layer 40 functionally corresponds to... Figure 1 The augmented reality navigation module 400 in the present invention is the human-machine interface (HMI) of the system. Figure 3 It clearly reveals the core position and working method of this layer from a perspective that focuses more on "roles" and "information flow".

[0130] In one specific embodiment, the core of the interaction layer 40 is the doctor (operator), using their AR device as the physical carrier. This layer is a key hub connecting machine intelligence and human clinical experience, and its information interaction is reflected in the following aspects: Receiving "AR Visualization Guidance" from Processing Layer 20: As shown by the arrows in the figure, Interaction Layer 40 receives generated virtual information from the "AR Navigation Module" of Processing Layer 20. As described in Embodiment 1 above, this information includes highlighted scar boundaries and planned suture trajectories, which are precisely overlaid on the doctor's real-world view, providing the doctor with an intuitive and clear "surgical map" to assist in precise positioning and operational supervision. Sending "Manual Fine-tuning Instructions" to Processing Layer 20: This invention is not intended to completely replace the doctor, but rather to extend their capabilities. Figure 3 As shown, while observing AR guidance, doctors have the highest authority to intervene in the system strategy. Through gestures, voice, or force feedback handles, doctors can issue manual fine-tuning commands (e.g., adjusting the position of a suture anchor point). These commands are sent back to the data fusion module in processing layer 20, which recalculates and optimizes the suture strategy in real time, embodying the essence of human-machine collaboration. Receiving "alarm signals" from safety layer 50: In addition to regular navigation information, interaction layer 40 is also a key receiver for system risk warnings. When the alarm module in safety layer 50 determines that risk conditions are met, it immediately sends an alarm signal to interaction layer 40. This signal can be conveyed to the doctor through the AR device in the most conspicuous way (e.g., flashing red at the edge of the field of vision) and / or through sound (e.g., a buzzer alarm), ensuring that the doctor is aware of the risk and makes timely intervention decisions.

[0131] Through this multi-directional information interaction of "visual guidance - real-time control - risk warning", the interaction layer 40 ensures that doctors are always in a central and dominant position in the entire intelligent surgical process.

[0132] Security layer 50, please continue reading Figure 3 Safety layer 50 is the final safety defense and core of risk control in the system of this invention. It is logically independent of the conventional closed-loop feedback control process and is designed to provide a highest priority, redundant safety guarantee for the entire surgical procedure.

[0133] In one specific embodiment, safety layer 50 mainly consists of an alarm module. This module is designed to react as quickly as possible to bring the system to a safe state when a serious or unexpected anomaly is detected. Its information interaction and operation are as follows: Receiving "abnormal signals" from processing layer 20: As shown by the arrows in the figure, the alarm module of safety layer 50 continuously receives signals marked as "abnormal" from the data fusion module of processing layer 20. These abnormal signals are not merely conventional mechanical data exceeding limits, but can also be more complex risks determined by AI image analysis logic. For example, in a preferred embodiment, processing module 200 performs dual verification on real-time tension data and tissue thickness data obtained through image analysis using a pre-trained anomaly detection model. When both conditions are met simultaneously—tension > 1.5N and tissue thickness > 8mm (far exceeding normal muscle layer thickness)—the system determines that this is a serious risk that conventional feedback adjustments may not be able to handle and immediately sends this as an "abnormal signal" to safety layer 50. Sending the highest priority instructions to multiple levels: Once the alarm module receives an abnormal signal and confirms the risk, it sends instructions to other levels in parallel within a very short time (e.g., within 0.2 seconds): Sending an "alarm signal" to the interaction layer 40: This uses the most conspicuous visual and auditory methods (such as global red flashing in the AR field of view and a high-decibel buzzer alarm) to convey the highest level of alarm to the doctor. Sending a "stop instruction" to the execution layer 30: This instruction has the highest priority and will immediately forcibly lock the robotic arm joints and cut off the power to the suture device to prevent any erroneous movements that could worsen the injury. Simultaneously, the module also records anomaly logs, including the time the alarm was triggered, data snapshots, and other information, for postoperative review and analysis.

[0134] In this way, safety layer 50 provides the entire system with a "circuit breaker mechanism" that is independent of the regular operating procedures, ensuring that the system remains safe and controllable even in the most extreme and unexpected emergencies, and maximizing the safety of patients.

[0135] In summary, Figure 3 From a layered architecture perspective, the complex logical relationships and information flows within the stitching and reinforcement system of this invention are revealed in greater depth. Specifically, data layer 10 provides comprehensive static and dynamic input to the system; processing layer 20, acting as the system's "intelligent brain," completes the entire process from data analysis to intelligent decision-making; execution layer 30 and interaction layer 40, as physical and information outputs, respectively achieve precise physical operations and efficient human-machine collaboration; and the logically independent security layer 50 provides the highest priority security guarantee for the entire system. It is this precise division of labor and seamless collaboration among these five layers that together constitute the stable, efficient, and secure closed-loop control system of this invention.

[0136] Preferred embodiment: Intelligent suturing experiment of a porcine scarred uterus model To verify the effectiveness, safety, and advanced nature of the suture reinforcement system provided by this invention in biological tissues in near-real clinical settings, this animal experiment was conducted.

[0137] Experimental Background and Preparation: This experiment selected three healthy experimental pigs weighing 40-50 kg that had successfully undergone cesarean section to construct a uterine scar model, in order to simulate the complex surgical scenario of cervical scar repair after multiple cesarean sections in humans. Before the surgery, thorough preoperative preparation and data collection were carried out, including: Preoperative data acquisition: Each experimental pig underwent magnetic resonance imaging (MRI) scans to obtain a three-dimensional model (1 mm slice thickness) of its uterine scar area. Specifically, the average elastic modulus of the scar tissue was measured and recorded using elastography sequences, with an average value of approximately 280 kPa. Hardware calibration: To ensure surgical precision, the augmented reality navigation module 400 was spatially calibrated using the NDI Polaris infrared positioning system, achieving a navigation accuracy of ±0.2 mm. Simultaneously, the UR3e robotic arm in the mechanical execution module 300 underwent no-load positioning error testing, confirming that its repeatability error was less than 0.3 mm, meeting the surgical requirements.

[0138] Intraoperative intelligent suturing operation and verification: After anesthesia and establishment of pneumoperitoneum, the suturing enhancement system of this invention was deployed and activated to begin intelligent suturing operations. The entire process clearly demonstrates the collaborative work of the various modules of the system: Data acquisition and real-time analysis: The data acquisition module 100 begins operation. The laparoscopy transmits images of the uterine scar area to the processing module 200 in real time. Simultaneously, the intelligent suture needle, integrated with a sensor array, monitors the tissue tension at the suture point in real time when in contact with the tissue. Data shows that in a typical scar area, the tension fluctuates between 1.2-1.8N, and the system's safety threshold is set at 1.5N. After receiving the image data, the processing module 200's internal deep learning model immediately performs real-time segmentation, accurately delineating the boundaries of the scar tissue. The segmentation accuracy was evaluated, with a Dice coefficient reaching 0.91.

[0139] Generation and execution of individualized suturing strategies: Path planning: Based on the preoperative 3D model, the processing module 200 plans a suturing trajectory containing 10 suturing anchor points spaced 3-4 mm apart, effectively avoiding the uterine artery (safe distance ≥ 3 mm). Dynamic speed adjustment: Based on the preoperatively measured average elastic modulus (280 kPa), the processing module 200 automatically sets the firing speed of the intelligent suture device to 15.2 mm / s (calculated using formula 12 + 0.04 * (280 - 200)) using its internal dynamic speed adjustment algorithm. Guidance and execution: The planned suturing trajectory is displayed clearly and in real-time in the operator's AR glasses field of view via the augmented reality navigation module 400. The mechanical execution module 300 begins precise suturing based on the trajectory and the set speed and depth parameters. Measurements show that the actual suturing positioning error is only 0.25 mm. During the operation, the surgeon fine-tuned the position of one of the anchor points using the force feedback handle. Upon receiving the instruction, the system automatically and smoothly updated the overall suture trajectory, demonstrating excellent human-machine collaboration. A case study verifying the risk closed-loop feedback mechanism: During the fifth stitch, a critical scenario occurred: the intelligent suture needle detected a sudden increase in tissue tension to 1.6N for 20ms, meeting the system's preset risk conditions. The processing module 200 immediately reacted, issuing an adjustment instruction to the mechanical execution module 300 within 0.2 seconds, automatically reducing the suture depth by 0.5mm. As the depth decreased, the detected tension quickly returned to the safe range of 1.3N, and the system then continued the suture operation. This process perfectly verified the speed, accuracy, and effectiveness of the closed-loop feedback mechanism of this invention.

[0140] After the experiment, analysis of the data and results showed that the system of the present invention exhibited significant beneficial effects: compared with the control group using traditional manual suturing, the average suturing time of the system of the present invention was reduced by 35%. Biomechanical analysis of the postoperative sutured area showed that the stress unevenness rate (an indicator of the degree of local stress concentration) was reduced from 32% in the traditional method to 12%. Comprehensive evaluation indicated that the system of the present invention reduced the risk of tissue tearing due to improper tension by approximately 60%.

[0141] This experimental example fully demonstrates that the suture enhancement system provided by this invention can work stably, efficiently, and safely in real and complex biological tissue environments, and its effect on improving suture accuracy and reducing surgical risks is significant and quantifiable.

[0142] Preferred embodiment: Verification of complex scar suturing using a silicone tissue model.

[0143] To further test the performance robustness and intelligent adaptation capability of the suture reinforcement system of this invention under various extreme and complex conditions, a tissue model experiment was specially designed and conducted.

[0144] Experimental Background and Preparation: This experiment did not use biological tissue, but instead used a specially customized silicone tissue model capable of simulating the most demanding surgical scenarios. This model, through precise material formulation and molding processes, integrates simulated scar areas with varying degrees of fibrosis. Its key parameters are set as follows: Elastic modulus: varying between 200 kPa and 500 kPa to simulate scar tissue ranging from ordinary scars to highly fibrotic and rigid scar tissue; Thickness: varying between 5 mm and 10 mm to simulate extreme cases of uneven tissue thickness caused by scar hyperplasia or adhesion. The purpose of this experiment is to quantitatively and reproducibly verify the performance limits and reliability of the core technologies of this invention's system under these controllable and standardized extreme conditions.

[0145] Core Technology Application and Verification, (I) Verification of Preoperative Planning Capability Before the experiment began, the computer-aided design (CAD) data of the silicone model was first imported into the processing module 200. The AI ​​algorithm in the processing module 200 can pre-generate three different suturing strategy schemes based on these geometric and material data, corresponding to scenarios such as "conventional scar", "highly fibrotic scar" and "excessively thick scar". A particular verification point is that the AI-generated strategy can adaptively adjust the precision of suturing. For example, in the area with an elastic modulus of 200 kPa, the preset anchoring points are 8; while in the area with a higher hardness of 500 kPa, the system automatically increases the density of anchoring points to 12, to ensure more reliable fixation in harder tissues. This proves that the system has the ability to perform refined planning based on prior knowledge. (II) Verification of Intraoperative Dynamic Adaptation Capability, Image Segmentation: In order to clearly identify the scar boundaries on the model, we sprayed fluorescent markers on its surface. In the experiment, the system's deep learning model accurately identified these markers, achieving a Dice coefficient of 0.93 for scar boundary segmentation, approximately 1% higher than in real biological tissue, demonstrating the algorithm's superiority. Tension monitoring and dynamic speed adjustment: When the intelligent suture needle sutured in a highly fibrotic area with an elastic modulus of 500 kPa, the sensor monitored tissue tension between 1.5 and 2.0 N, stably triggering the system's low-speed protection mode. The suture needle's firing speed automatically decreased from the conventional 18 mm / s to 12 mm / s. Under this crucial adaptation, the failure rate of suture formation in this area significantly decreased from 17% to 9%. Human-machine collaboration: In the experiment, the operator (doctor) actively selected and switched to the preset "thick scar mode" through the interactive interface. Upon receiving this instruction, the system immediately responded: automatically increasing the suture depth to 5 mm and displaying the guide line for layered suturing in the AR view, perfectly demonstrating the seamless integration of human-machine commands and system strategies. The safety mechanism stress test was conducted to assess the system's risk assessment and emergency response capabilities. An extreme anomaly scenario of "excessively thick tissue" was deliberately simulated, where the suture device attempted to clamp a 9mm thick silicone tube. At that instant, the system's safety mechanism was successfully triggered: the processing module 200 detected the anomaly within 0.2 seconds and immediately sent a stop command to the mechanical execution module 300.

[0146] Simultaneously, the augmented reality navigation module 400 immediately displays a prompt in the operator's field of vision: "It is recommended to stitch in two layers." After the operator manually confirms the suggestion with gestures, the system automatically and instantly regenerates a new stitching trajectory containing 5 additional intermediate anchor points.

[0147] The data from this experiment are highly convincing: when faced with high-hardness scars with an elastic modulus as high as 500 kPa, the suturing success rate of the system of this invention reached 92%, while the success rate of the traditional fixed-parameter method was only 65%. In the entire safety mechanism stress test, the system's abnormal state detection accuracy reached 100%, with no missed or false detections.

[0148] This experimental example fully demonstrates that the suture reinforcement system provided by the present invention not only performs excellently in conventional biological tissues, but also maintains stable and reliable intelligent adaptation capabilities, human-machine collaboration capabilities, and safety redundancy mechanisms when facing various foreseeable extreme and complex scenarios.

[0149] This application provides an intelligent suturing system and method for cesarean section scarred uterus. Through a collaborative architecture of "perception-decision-execution," it achieves end-to-end intelligent intervention for suturing complex scarred uteruses. First, by integrating multimodal sensors onto the suture needle and deeply fusing its real-time biomechanical sensing data with image information segmented by a deep learning model, the operator's "vision" and "touch" are transformed into unified, high-dimensional machine cognition, fundamentally breaking down "information silos" and providing a high-quality data foundation for generating global and precise suturing strategies. Second, this system utilizes a dynamic speed adjustment algorithm based on individualized biomechanical parameters (such as elastic modulus) and uses real-time biomechanical sensing data as input for closed-loop feedback, constructing a decision-making closed loop that combines foresight and real-time performance. This transforms suturing decisions from a passive, doctor-experience-dependent "art" to an active, real-time data-driven "science." Furthermore, by introducing a parallel dual-verification risk monitoring mechanism, it achieves second-level detection and immediate response to potential tissue tearing and other risks, transforming the traditional model of relying on doctors for post-operative remediation into a new model of proactive prevention and immediate intervention. Through the synergistic effect of the above-mentioned key technical links, the present invention system overcomes the limitations of traditional suturing methods, realizes comprehensive perception, precise planning, adaptive adjustment and intelligent risk control of the suturing process, and significantly improves the safety and effectiveness of high-risk gynecological surgeries such as multiple cesarean sections.

[0150] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0151] The terms "an embodiment," "embodiment," or "one or more embodiments" as used herein mean that a particular feature, structure, or characteristic described in connection with an embodiment is included in at least one embodiment of this application. Furthermore, please note that the examples of the phrase "in one embodiment" do not necessarily all refer to the same embodiment.

[0152] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0153] In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In a unit claim enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0154] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. An intelligent suturing system for uterine scars from cesarean section, characterized in that, include: The data acquisition module includes: an image acquisition device configured to acquire image data of the surgical area in real time; and a smart suture needle with a pressure sensor and / or strain sensor integrated on its body, configured to acquire multimodal mechanical sensing data in real time when interacting with uterine tissue. The processing module, coupled to the data acquisition module, is configured to: identify and segment the boundary of scar tissue in real time using a deep learning model based on the image data; and dynamically generate an individualized suturing strategy that includes suture trajectory, suture depth, and / or suture firing speed parameters by fusing the boundary information of the scar tissue with the mechanical sensing data. A mechanical execution module, coupled to the processing module, is configured to receive the individualized suturing strategy to drive its included robotic arm and suture device to perform suturing operations. An augmented reality navigation module, coupled to the processing module, is configured to overlay the boundaries of the scar tissue and the suture trajectory in the form of virtual information onto the real-time view of the surgical area for the operator to observe. The processing module is further configured to continuously monitor the mechanical sensing data during the suturing operation performed by the mechanical execution module, and to send control commands to the mechanical execution module to automatically and in real time adjust the suturing operation when the mechanical sensing data meets one or more preset risk conditions.

2. The system according to claim 1, characterized in that, The structure of the deep learning model includes: An encoder employs a multi-scale dilation attention mechanism to capture multi-scale features of scar tissue through different dilation rates in parallel branches, and uses a sliding window attention mechanism to reduce computational complexity; and A decoder employs a multi-feature channel attention fusion mechanism, combining global and local features to enhance and output the scar tissue boundary.

3. The system according to claim 1, characterized in that, When generating the individualized suturing strategy, the processing module is also configured to: Estimate the elastic modulus Es of the scar tissue based on preoperative elastography data or intraoperative stress-strain curves; and The stitching firing speed is determined based on a preset piecewise function relationship related to the elastic modulus Es, wherein the stitching firing speed decreases as the elastic modulus Es increases.

4. The system according to any one of claims 1, characterized in that, The processing module is configured to determine whether one or more preset risk conditions are met; wherein, the one or more preset risk conditions include at least: The pressure value of the pressure sensor continuously exceeds a preset first pressure threshold within a preset time period, and the strain value of the strain sensor exceeds a preset second strain threshold.

5. The system according to any one of claims 1, characterized in that, The processing module is further configured to, based on preoperative medical imaging data, use the Dijkstra algorithm or a variant thereof to plan an initial suture path of the suture trajectory while avoiding areas with dense blood vessels, and perform B-spline interpolation smoothing on the initial suture path.

6. A method for strengthening the suture of a cesarean section scar uterus through intelligent suturing, characterized in that, Includes the following steps: The image acquisition device acquires real-time image data of the surgical area, and the sensor integrated on the smart suture needle collects multimodal mechanical sensing data in real time when interacting with uterine tissue. The image data is processed using a deep learning model to segment the boundaries of scar tissue in real time; By integrating the boundary information of the scar tissue with the mechanical sensing data, an individualized suturing strategy is dynamically generated, which includes parameters such as suture trajectory, suture depth, and / or suture firing speed. Using augmented reality devices, the boundaries of the scar tissue and the suture trajectory are superimposed on the real view of the surgical area in the form of virtual information; The mechanical actuation module controls the suturing operation according to the individualized suturing strategy; as well as During the suturing operation, the mechanical sensing data is continuously monitored, and the suturing operation is automatically and in real time adjusted when the mechanical sensing data is detected to meet one or more preset risk conditions.

7. The method according to claim 6, characterized in that, The step of processing the image data using a deep learning model includes: Multi-scale features of the image data are extracted using an encoder with both multi-scale dilation attention and sliding window attention mechanisms; and The boundary of the scar tissue is enhanced and output through a decoder with a multi-feature channel attention fusion mechanism.

8. The method according to claim 6, characterized in that, The step of generating the individualized suturing strategy further includes: Estimate the elastic modulus Es of the scar tissue; The stitching firing speed is determined based on a preset piecewise function relationship related to the elastic modulus Es.

9. The method according to claim 6, characterized in that, The preset risk conditions include at least the following: The pressure value derived from the mechanical sensing data is continuously greater than a first pressure threshold within a preset time period, and the strain value in the mechanical sensing data is simultaneously greater than a second strain threshold.

10. The method according to any one of claims 6 to 9, characterized in that, Prior to the step of generating the individualized suturing strategy, the method further includes: Based on preoperative medical imaging data, the Dijkstra algorithm was used to plan an initial suture path that avoids areas with dense blood vessels; And the initial suture path is smoothed by B-spline interpolation.

Images