A Visual Analysis-Based Intelligent Monitoring Method for Electric Operating Tables

By using visual analysis and multi-dimensional monitoring links, the problem of real-time monitoring of micro-drift and motion trajectory deviation of electric operating tables during surgery was solved, realizing multi-dimensional quantitative assessment and early warning of electric operating tables, and improving the safety and reliability of operating tables.

CN123101352APending Publication Date: 2026-11-17JINING CENT FOR FOOD & DRUG CONTROLJINING CENT FOR ADVERSE DRUG REACTION & DRUG ABUSE MONITORING
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611344098.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-01
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing electric operating table monitoring solutions struggle to capture slow drifts in the bed's position and movement trajectory deviations in real time during surgery, and lack quantitative assessment methods, leading to reduced safety and reliability in the use of the operating table.

Method used

A vision-based intelligent monitoring method is adopted. By real-time detection of the key point set of the operating table and calculation of the pose vector, combined with depth point cloud segmentation and inertial measurement unit, a multi-dimensional monitoring link is constructed to achieve comprehensive quantitative assessment and early warning of micro-drift, motion trajectory deviation and overturning risk.

Benefits of technology

It enables multi-dimensional real-time automated monitoring of electric operating tables, which can identify micro-drift trends at an early stage, prevent patient injuries caused by loss of control of movement, and quantify the risk of tipping over in real time, thereby improving the safety and reliability of the operating table.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123101352A_ABST
    Figure CN123101352A_ABST
Patent Text Reader

Abstract

This application discloses a visual analysis-based intelligent monitoring method for electric operating tables, relating to the field of visual intelligent monitoring technology. The method includes: performing target detection on enhanced images of the operating table to obtain a pose vector; predicting the cumulative amount of body position deviation and outputting a micro-drift risk index; determining a local window during body position transitions, generating a three-dimensional motion trajectory, comparing the real-time three-dimensional motion trajectory frame-by-frame with a safe motion trajectory template, and outputting a trajectory deviation risk index; segmenting the patient's surface point cloud, calculating the patient's overall center of gravity position, calculating the dynamic stability margin, and outputting a tipping risk index; estimating key points on the patient's surface, calculating slippage relative to the bed surface, and outputting a slippage risk index; generating an intraoperative monitoring and analysis report; and achieving a comprehensive intelligent assessment of the intraoperative safety risks of the electric operating table through multi-dimensional real-time monitoring and comprehensive scoring of bed pose micro-drift, motion trajectory deviation, center of gravity tipping risk, and patient slippage.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of visual intelligent monitoring technology, and in particular to an intelligent monitoring method for an electric operating table based on visual analysis. Background Technology

[0002] Electric operating tables are an indispensable core piece of equipment in the operating room, and their mechanical stability and electrical safety directly affect the success or failure of the surgery.

[0003] Currently, the monitoring of electric operating tables mainly involves the following technical approaches: First, indirect assessment of the table's condition through manual observation by medical staff and intraoperative monitoring of the patient's vital signs. This method relies on the clinical experience of medical staff, cannot achieve quantitative or continuous monitoring, and requires staff to maintain close attention to the table's condition during surgery due to their high concentration on the procedure itself. Second, monitoring single-parameter thresholds using sensors integrated into the electric operating table, such as detecting motor current overload or exceeding travel limits. This method cannot capture slowly developing abnormal trends or assess the synergistic effects of multiple parameters. Third, a periodic maintenance system, which involves mechanical and electrical inspections of the operating table at fixed intervals according to the equipment's instruction manual. This method is a periodic preventative measure and cannot cover sudden abnormalities occurring between maintenance intervals, nor can it proactively identify real-time risks during surgery.

[0004] Existing electric operating table monitoring solutions struggle to capture the slow drift of the patient's position during surgery in real time, and lack quantitative assessment methods for motion trajectory deviations during position adjustments. These deficiencies prevent multi-dimensional, real-time, automated monitoring of the electric operating table's intraoperative status, reducing the safety and reliability of the operating table's use. Summary of the Invention

[0005] This application provides a visual analysis-based intelligent monitoring method for electric operating tables, which solves the problem in the prior art that it is impossible to perform multi-dimensional real-time correlation monitoring of intraoperative micro-drift and movement trajectory deviation of electric operating tables. It achieves the technical effect of comprehensive quantitative assessment of safety risks such as intraoperative micro-drift and movement trajectory deviation of operating tables.

[0006] This application provides a visual analysis-based intelligent monitoring method for electric operating tables, including: S1: Preprocess the global illumination of each operating table image to obtain an enhanced image sequence; perform real-time target detection on each image in the enhanced image sequence to identify the key point set of the operating table and obtain the pose vector of each key point; S2: Based on the pitch and roll angle time series of the entire bed surface and backboard in the pose vector, calculate the drift velocity and drift acceleration, predict the cumulative body position deviation at multiple key time points in the future, and output the micro-drift risk index. S3: During the body position transformation process, the pose vector is obtained and several local windows are determined. A three-dimensional motion trajectory is generated in each local window. The real-time three-dimensional motion trajectory is compared with the safe motion trajectory template frame by frame. The deviation index set is calculated and the trajectory deviation risk index is output. S4: Use depth point cloud to segment the patient's body surface point cloud and calculate the patient's overall center of gravity position; determine the bed support polygon and calculate the shortest distance from the center of gravity projection point to each side of the support polygon; define the ratio of the shortest distance to the radius of the circumcircle of the bed base as the static stability margin; calculate the inertial force reduction to obtain the dynamic stability margin and output the overturning risk index. S5: Estimate key points on the patient's body surface using human posture, calculate slippage relative to the bed surface, and output slippage risk index by combining BMI adaptive threshold; S6: Based on steps S1 to S5, construct a multidimensional monitoring index set and generate an intraoperative monitoring and analysis report.

[0007] Furthermore, the method also includes: S51: Construct a bed-patient relative pose correlation model to distinguish between passive patient displacement caused by active bed movement and abnormal patient slippage when the bed is stationary; including: calculating the correlation coefficient between the change in bed pose and the change in patient key point displacement. A correlation threshold range is set, including an upper correlation threshold and a lower correlation threshold. When the correlation coefficient is greater than the upper correlation threshold, it is determined to be passive displacement caused by active bed movement, which is a normal phenomenon. When the correlation coefficient is less than the lower correlation threshold and the patient's key point displacement is continuously non-zero, it indicates that the patient has voluntary slippage while the bed is stationary, which is an abnormal phenomenon.

[0008] Furthermore, the method also includes: S52: Calculate the transfer function between the change in bed pitch angle and the patient's shoulder displacement, establishing a spatial mapping relationship, i.e., the amount of patient displacement caused by a unit angle change; the transfer function has different values ​​under different surgical positions. Based on the transfer function, the current change in bed pitch angle is weighted to obtain the expected displacement of the patient's shoulder driven only by bed movement; define the difference threshold as the deviation tolerance between the measured patient displacement and the expected displacement. When the absolute value of the difference exceeds the difference threshold, there is an additional slip driving force, triggering an association anomaly warning; output the association anomaly index, calculated as the ratio of the absolute value of the difference between the measured patient displacement and the predicted displacement to the difference threshold.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages: By analyzing the time series of pitch and roll angles of the entire bed surface and backrest, and performing drift velocity and drift acceleration analysis, a linear extrapolation prediction model is constructed to achieve early identification of micro-drift trends and prediction of cumulative drift. Compared to existing technologies that can only detect obvious angular deviations that have already occurred, this invention can issue an early warning before the drift amount reaches the clinically dangerous threshold, allowing time for intervention.

[0010] By tracking the three-dimensional motion trajectory in real time during body position transitions and comparing the measured trajectory with a safe motion trajectory template frame by frame, a multi-dimensional quantitative assessment of speed deviation, path deviation, and positioning accuracy deviation is achieved. When the deviation exceeds the limit, a graded warning or motion pause is automatically triggered, effectively preventing patient injury caused by loss of movement control during body position adjustments.

[0011] By segmenting the deep point cloud to obtain the patient's surface point cloud, the overall center of gravity position of the patient is calculated and combined with the bed support polygon and inertial force correction to obtain the dynamic stability margin, thereby realizing the real-time quantitative assessment of the risk of tipping over during the body position adjustment process and overcoming the limitations of existing solutions that rely solely on static center of gravity judgment. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of a visual analysis-based intelligent monitoring method for an electric operating table in an embodiment of the present invention. Detailed Implementation

[0013] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0014] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0015] The intelligent monitoring method for electric operating tables involved in this invention covers two working states of electric operating tables: periodic routine monitoring in non-surgical state and real-time intelligent monitoring during surgery. The two states share the same multimodal visual perception hardware architecture and pose estimation benchmark, but each has its own emphasis in monitoring strategy, data acquisition frequency and early warning response mechanism.

[0016] In non-surgical situations, the electric operating table is in idle standby, between surgeries, or during routine maintenance, when the patient is not on the table and the table is unloaded. It automatically switches to a periodic routine monitoring mode, performing standardized motion sequence tests according to a preset testing plan. These tests include independent lifting of each part, overall tilting, and folding of the headboard and legboards. The vision system records the actual movement trajectory and final positioning accuracy of each part in real time, comparing it with the factory-calibrated baseline values ​​to calculate repeatability and absolute positioning accuracy. Simultaneously, when the table is stationary and locked, the vision system continuously monitors the static holding capability of each part, judging whether there are hidden leaks in the hydraulic locking system or mechanical braking device by the amplitude of posture fluctuations within a preset time period. The results of each routine test are compared longitudinally with the historical test data of the table, generating performance degradation curves such as the trend of positioning accuracy over time and the trend of static holding capability over the number of uses. When a performance indicator degrades to near the maintenance threshold, a preventative maintenance work order is automatically generated. All test data in non-surgical situations is stored in the operating table's electronic health record, providing data support for equipment replacement and decommissioning decisions.

[0017] During surgery, the safety risks of patients under anesthesia on an electric operating table are far greater than those of patients not undergoing surgery. Potential adverse events include slow positional drift due to hydraulic system leaks, deviations from the intended trajectory due to motor or transmission mechanism malfunctions, tipping risks due to changes in the center of gravity during significant position adjustments, and relative slippage along the bed surface caused by the patient's weight. These risks are characterized by their suddenness, cumulative nature, and high nonlinearity. Traditional monitoring methods rely primarily on manual observation by medical staff and passive postoperative reporting, lacking real-time, quantitative, and automated risk identification capabilities.

[0018] Existing monitoring solutions for electric operating tables mostly focus on threshold alarms for a single parameter, failing to correlate the kinematic parameters of the bed itself with the biomechanical response of the patient's body surface, and also making it difficult to capture the slowly accumulating micro-drift trend.

[0019] To address the aforementioned issues, this application proposes a visual analysis-based intelligent monitoring method for electric operating tables. This method, based on a shared set of multimodal visual perception hardware, constructs a multi-layered monitoring chain for real-time monitoring needs during surgery. This chain includes bed pose estimation, micro-drift accumulation prediction, motion trajectory tracking, tilt risk assessment, and patient-bed relative slippage detection. Through multi-source data fusion and comprehensive risk scoring, tiered early warnings are achieved, ultimately forming a full-lifecycle intelligent monitoring system covering real-time monitoring during surgery and periodic non-surgical monitoring.

[0020] This application aims to monitor an electric operating table in real time during surgery. It requires the pre-deployment of several sensors. For example, six RGB cameras and three depth cameras are deployed at the four corners of the operating room ceiling and on both side walls. The RGB cameras have a resolution of at least 1920×1080 pixels and a frame rate of 30 frames per second. The depth cameras employ time-of-flight or structured light principles, with a frame rate of 15 frames per second and a depth measurement accuracy of at least 1 millimeter. Eight miniature inertial measurement units are deployed at the four corners of the operating table surface and at the hinge points of the headboard, backboard, seat, and legboards. Each unit incorporates a three-axis accelerometer and a three-axis gyroscope, with an output frequency of 200 Hz.

[0021] All sensors are timestamped using a hardware synchronization signal generator. Zhang's calibration method is used to calibrate all nine cameras, obtaining the intrinsic and extrinsic parameter matrices for each camera. All cameras are then aligned to the operating room world coordinate system with the center of the operating room floor as the origin. The inertial measurement unit determines the direction of gravity using accelerometer readings during the initial stationary phase, completing the initial attitude alignment with the operating room world coordinate system.

[0022] Example 1: As Figure 1 As shown, a method for intelligent monitoring of an electric operating table based on visual analysis is disclosed, the method comprising: S1: Preprocess the global illumination of each operating table image to obtain an enhanced image sequence; perform real-time target detection on each image in the enhanced image sequence to identify the key point set of the operating table and obtain the pose vector of each key point.

[0023] The strong local light from the operating room's shadowless lamps, the metallic reflection from surgical instruments, and the dynamic shadows cast by medical staff can severely interfere with the detection accuracy of operating table feature points and patient surface landmarks in RGB images.

[0024] The global illumination state of each image is evaluated and preprocessed to obtain an enhanced image sequence. Specifically, the global illumination state of each image is evaluated using the following method: For each frame of RGB image, three feature values ​​are calculated: mean luminance, standard deviation luminance, and kurtosis luminance distribution. The mean luminance is the arithmetic mean of the luminance values ​​of all pixels in the image, reflecting the overall illumination level; the standard deviation luminance reflects the uniformity of the image illumination; and the kurtosis luminance distribution reflects the concentration of the illumination distribution. These three feature values ​​constitute an illumination state vector, used to characterize the illumination conditions of the current frame.

[0025] Based on the illumination state vector, the camera's exposure time and gain are automatically adjusted to stabilize the average image brightness within a preset target brightness range. When the average brightness exceeds the upper limit of the target range, the exposure time is reduced proportionally; when the average brightness falls below the lower limit of the target range, the exposure time or gain is increased proportionally. The step size of the exposure adjustment is proportional to the deviation of the current brightness from the target range, avoiding drastic oscillations in the exposure parameters. Proportional increases or decreases in automatically adjusting camera parameters are well-known technical practices, and this application does not impose specific limitations here.

[0026] The RGB image is converted to the HSV color space, and the hue, saturation, and lightness channels are separated. In the lightness channel, an adaptive thresholding method is used to identify local bright areas. Pixels with lightness values ​​greater than the global lightness mean plus twice the standard deviation are marked as bright areas. For pixels within bright areas, interpolation repair is performed using the hue and saturation information of surrounding non-bright areas. Specifically, a neighborhood window is taken centered on the bright pixel, and the median hue and saturation values ​​of non-bright pixels within the window are calculated. These two medians replace the hue and saturation values ​​of the bright pixel, while the lightness value is compressed to within 1.2 times the global lightness mean. This effectively suppresses overexposure in bright areas while preserving color information.

[0027] The system continuously monitors the rate of change of the average image brightness between adjacent frames. When the absolute value of the rate of change exceeds a preset threshold, a sudden change in illumination is identified. At this point, a shadow compensation mechanism is activated: the current frame is matched with the previous frame's brightness histogram to establish a brightness mapping function. The brightness of each pixel in the current frame is converted to compensated brightness through this function, ensuring that the brightness histogram of the compensated image is consistent with that of the previous frame. Simultaneously, temporal brightness filtering is applied to the same spatial location across multiple consecutive frames to suppress illumination jitter caused by random noise. After this image preprocessing, an enhanced RGB image sequence is output, where the inter-frame fluctuation of the average brightness is controlled within a preset range, pixel information in bright areas is effectively restored, and contrast in shadow areas is improved.

[0028] The key point set includes at least the headboard, backboard, seat board, legboard, and the entire bed surface; for each key point, two-dimensional image feature points are extracted, and the two-dimensional image feature points are back-projected into the operating room world coordinate system to obtain the corresponding three-dimensional spatial coordinates; the two-dimensional image feature points are the coordinates of the four corner points and the center point of the key point; combined with the depth point cloud output by the depth camera at the same time, the two-dimensional image feature points are back-projected into the operating room world coordinate system.

[0029] For each keypoint, a six-degree-of-freedom pose vector is defined, which includes the X, Y, and Z coordinates of the site center point in the operating room world coordinate system, as well as the roll angle (rotation angle about the X-axis), pitch angle (rotation angle about the Y-axis), and yaw angle (rotation angle about the Z-axis). These three angles can also be collectively referred to as attitude angles.

[0030] In this embodiment, the yaw angle is used as an auxiliary monitoring parameter for tracking the bed's position in non-surgical states. Specifically, during the downtime between surgeries, when the operating table is moved by casters (e.g., from one side of the operating room to the other, or transferred between different operating rooms), the yaw angle reflects the bed's orientation change in the horizontal plane. Combined with the X and Y coordinates, the movement trajectory of the bed within the operating room can be completely tracked, allowing for analysis of the bed's stability during transport. When a rapid change occurs within a short period (e.g., angular velocity exceeding 10 degrees per second), it indicates that the bed has made a sharp turn or sideslip during movement, potentially causing impact loads on the internal mechanical structure. This event is recorded in the bed's electronic health record as reference data for cumulative fatigue analysis of the mechanical structure. During surgery, the bed is locked, and the yaw angle remains constant, not participating in intraoperative risk assessment. However, it is still necessary to record the yaw angle to ensure data integrity.

[0031] The frequency of visual positioning is limited by the image acquisition frame rate. Between adjacent visual positioning frames, pose prediction is performed using high-frequency acceleration and angular velocity data provided by the inertial measurement unit (IMU). Specifically, an extended Kalman filter framework is used for fusion: the prediction equation uses the angular velocity and acceleration from the IMU for pose recursion; the update equation corrects the state vector using the visual positioning result as the observation value when visual positioning data arrives. The frequency of the fused pose data is synchronized with that of the IMU, achieving a pose accuracy better than 1 mm in the position dimension and better than 0.05 degrees in the angle dimension.

[0032] S2: Based on the pitch and roll angle time series of the entire bed surface and backboard in the pose vector, calculate the drift velocity and drift acceleration, construct a linear extrapolation prediction model, predict the cumulative body position deviation at multiple key time points in the future, and output the micro-drift risk index.

[0033] In some embodiments, when the cumulative predicted offset at any key time point exceeds a preset safety boundary, a micro-drift risk is determined; the key time points include at least three time points and the time point intervals shall not be less than three minutes; the cumulative body position offset includes the cumulative pitch angle offset and the cumulative roll angle offset.

[0034] Internal leakage in the hydraulic system of an electric operating table or failure of the mechanical locking mechanism typically manifests as extremely slow positional drift, with a typical drift rate between 0.01 and 0.1 degrees per minute. This level of drift is completely undetectable within a single frame or a short time window, but can accumulate to a significant shift of several degrees over thirty minutes, potentially causing serious consequences in delicate surgeries such as neurosurgery. Using the time series of pitch and roll angles of the entire table surface and backrest from the output fused pose data as input, a cumulative effect analysis of micro-positional drift is performed.

[0035] In this embodiment, the drift speed is defined as the rate of change of body position angle within a preset time window, in degrees per minute; the drift acceleration is defined as the rate of change of drift speed within a preset time window, in degrees per square minute.

[0036] Based on the current drift velocity and drift acceleration, a linear extrapolation prediction model is constructed, and a set of key time points are set to make a step-by-step prediction of the drift accumulation at different time scales in the future, predicting the cumulative body position shift after the key time points.

[0037] Insufficient number of key time points cannot effectively cover drift trends and make it difficult to distinguish between short-term fluctuations and long-term systematic drift. The time interval between adjacent time points should be no less than three minutes. Since the micro-drift of the electric operating table is an extremely slow mechanical drift process, the natural change in body position angle within three minutes is usually only between 0.03 and 0.3 degrees. This change is significantly higher than the angle measurement noise of the visual-inertial fusion positioning system (better than 0.05 degrees), and can be reliably captured within the effective resolution range of the sensor. This ensures that there is a statistically significant signal difference between adjacent prediction points rather than being drowned out by noise. Secondly, the three-minute interval provides the surgical team with a basic time window from receiving the warning, assessing the situation, to responding to clinical decisions, making the warning have practical processing value.

[0038] Three or more prediction points provide a tiered predictive perspective encompassing short, medium, and long term. Short-term points capture immediate trends in drift and identify sudden accelerations in drift; medium-term points assess the persistence and stability of drift, determining whether it is a systemic failure rather than random fluctuation; long-term points assess the clinical risk of cumulative effects, determining whether a breach of safety boundaries is imminent. This combination ensures that risk assessment avoids both the high false alarm rates caused by random fluctuations within short time windows and the loss of timely warnings due to excessively long time windows.

[0039] Preferably, critical time points can be evenly distributed at equal intervals, or they can be set at non-equal intervals according to the surgical stage and the urgency of the risk. When the risk level is low, a uniform distribution of points with longer time intervals is used; when a continuous increase in drift velocity is detected, the time interval between critical time points is automatically shortened and the number of points is increased, providing more intensive prediction for high-risk periods. The specific values ​​of the points are determined in real time by the system based on the current drift velocity and drift acceleration, rather than by preset fixed values, thus achieving adaptive adaptation to different working conditions.

[0040] Based on the aforementioned key time points, a linear extrapolation model is used for each point to predict the cumulative body positional offset at that time point. The cumulative body positional offset includes both pitch angle offset and roll angle offset. The formulas for both are as follows: ;in, It represents the cumulative postural shift; t is the current time. Let be the drift velocity (degrees per minute) at the current time t. Let t be the drift acceleration at the current time t (degrees / minute²), and T be the predicted time length (minutes). The formula is based on the assumption of uniform acceleration motion and is applicable to the prediction of drift trends in the short to medium term (several minutes to tens of minutes). It can effectively describe the drift process of slow acceleration or deceleration.

[0041] Both the cumulative pitch offset and the cumulative roll offset are calculated according to the above formulas, and the corresponding time series can be used for the calculation.

[0042] When the cumulative prediction offset at any key time point exceeds the preset safety boundary, a micro-drift risk is identified. The micro-drift risk index is output, calculated using the following formula: Where T is the set of all key time points; It is a micro-drift risk index; These are the safety boundaries corresponding to the cumulative offset. Separate safety boundaries are set for the pitch and roll angles, with their values ​​set independently based on the type of surgery. For example, in neurosurgery, where pitch accuracy is extremely high, it is set to 0.5 degrees; the roll angle, typically adjusted less or to 0 degrees in neurosurgery, is set to 0.3 degrees. In orthopedic surgery, both pitch and roll safety boundaries are set to 1.0 degrees. For general surgeries, they are set to 1.5 degrees and 1.0 degrees respectively. A yellow warning is triggered when the micro-drift risk index exceeds 0.8, indicating a need to monitor changes in body position; an orange warning is triggered when the ratio reaches or exceeds 1.0, suggesting an inspection of the hydraulic system or mechanical locking mechanism.

[0043] The method further includes: S3: During the body position transformation process, the pose vector is obtained and several local windows are determined, and a three-dimensional motion trajectory is generated in each local window; the real-time three-dimensional motion trajectory is compared with the safe motion trajectory template frame by frame, the deviation index set is calculated, and the trajectory deviation risk index is output.

[0044] In some embodiments, the local window is the time period from the start time to the end time of the body position change. The start time is the moment when the rate of change of posture of any part of the bed exceeds a preset speed threshold (dynamically set according to the general design standards of the operating table and actual needs, without specific restrictions, such as the rate of change of pitch angle greater than 0.5 degrees per second), and the end time is the moment when the rate of change of posture of all parts is less than 0.1 degrees per second. Within the local window, the three-dimensional position coordinates and attitude angles of each part are recorded at fixed time intervals to form a three-dimensional motion trajectory.

[0045] In this embodiment, a safe motion trajectory template library is pre-built, and the real-time motion trajectory is compared with the corresponding template frame by frame. Different safe motion trajectory template libraries are pre-built for different surgical types and body position conversion requirements.

[0046] Specifically, standard position transition types are categorized based on the clinical use scenarios of the electric operating table. Each type has a clearly defined combination of starting position, target position, and motion axis. Under standard testing conditions, the operator controls the operating table to complete the standard actions for each type of position transition. The three-dimensional motion trajectory of each body part is synchronously acquired through the visual perception system deployed in this application. Each type of action is repeated multiple times to cover the normal fluctuation range. Spatiotemporal alignment is performed on multiple trajectories of the same type: temporal alignment uses a dynamic time warping algorithm to unify the duration standard of each trajectory, and spatial alignment unifies the trajectory to the local coordinate system of the operating table. This is existing technology and will not be elaborated upon in this application. After alignment, the mean and standard deviation of the position coordinates at each sampling time are calculated. The mean position constitutes the ideal position-time curve, and the mean position plus or minus twice the standard deviation constitutes the allowable deviation range of the path. The upper limit of speed is taken as 1.2 times the maximum speed of the taught trajectory, and the upper limit of acceleration is similarly calculated. Finally, the above parameters are combined into structured template entries and stored in the template library. The templates can be incrementally updated as actual operation data accumulates to maintain their timeliness.

[0047] Each safe movement trajectory template includes: a body position transition type identifier, the ideal position-time curve for each body part during the transition, an upper limit for speed, an upper limit for acceleration, and a permissible deviation range for the path. The upper limit for speed is set as follows: the rate of change of the backboard pitch angle should not exceed 3 degrees per second; the upper limit for acceleration is set as: the acceleration of the backboard pitch angle should not exceed 2 degrees per second; the permissible deviation range for the path is set as: positional deviation for each body part should not exceed 5 mm, and angular deviation should not exceed 0.5 degrees. These setting principles can be adjusted according to actual needs to ensure a smooth start and stop process and avoid patient swaying caused by sudden stops and starts. This application does not impose specific restrictions on these aspects.

[0048] The deviation index set includes speed deviation rate, path deviation rate, and positioning accuracy deviation rate; specifically, the speed deviation rate is the ratio of the difference between the real-time speed and the template speed to the upper limit of the speed; the path deviation rate is the ratio of the Euclidean distance between the real-time position and the ideal position of the template to the allowable deviation of the path under the current motion progress ratio; the positioning accuracy deviation rate is the ratio of the Euclidean distance between the final stopping position and the target position to the standard allowable value (set to 3 mm in this application, according to the industry general standard).

[0049] The determination of the path tolerance follows a dynamic adjustment principle, rather than a fixed value. Specifically, the current movement progress ratio is defined as the percentage of completed movement time out of the total planned movement time. The path tolerance is set in segments according to the movement progress ratio: in the initial stage (movement progress ratio less than 10%) and the final stage (movement progress ratio greater than 90%) of the position change, the bed is close to the patient or has approached the target position, at which point the positional accuracy requirement is the highest, and the tolerance is tightened to 3 mm; in the middle stage of the position change ([10%, 90%]), the bed is in motion, and the immediate requirement for positional accuracy is relatively relaxed, with the tolerance linearly transitioning from 3 mm to 8 mm and then back to 3 mm. The maximum value of 8 mm comes from the mechanical positioning redundancy range of the operating table in the non-precision stage. The above 3 mm value is based on the advance warning ratio of 60% for the repeatability accuracy requirement (5 mm) of the operating table in the national standard, while the 8 mm value is based on the upper limit of the natural swing amplitude of the operating table's mechanical structure during movement. Exceeding this amplitude indicates mechanical loosening or control abnormality. The permissible deviation in angle direction is uniformly set at 0.5 degrees. This value corresponds to the smallest meaningful deviation unit for adjusting the surgical position angle in clinical practice. Angle changes smaller than this value have a negligible impact on the surgical operating space.

[0050] The formula for calculating the path deviation rate is: ;in, It is the path deviation rate; The three-dimensional Euclidean distance between the real-time position and the ideal position of the template. This represents the allowable deviation of the path at the current motion progress ratio. This indicates that the real-time location deviates from the allowable range.

[0051] The formula for calculating the positioning accuracy deviation rate is: ;in, It is the positioning accuracy deviation rate; This is the three-dimensional Euclidean distance between the actual stopping position and the target position after the bed has come to a complete stop, expressed in millimeters. The denominator is 3 millimeters, based on 60% of the allowable value of 5 millimeters for repeatability of the operating table in the national standard. This serves as the warning trigger line, ensuring that intervention time is reserved before the deviation reaches the mechanical tolerance limit.

[0052] When any one of the three deviation indicators is greater than 1, a Level 1 warning is triggered, prompting attention to deviations in the current body position adjustment process through sound and light.

[0053] When the speed deviation rate is greater than 1 (i.e., the real-time speed exceeds the speed limit), or the path deviation is greater than the allowable path deviation, or the positioning accuracy deviation is greater than 3 mm, a first-level warning (audio-visual prompt) is triggered. When the speed deviation rate is greater than 1.3 (i.e., exceeding the upper limit by 30%) or the path deviation rate is greater than 1.5 (i.e., the path deviation exceeds 50% of the current allowable deviation), a second-level warning is triggered, and a pause command is automatically sent to the operating table control system.

[0054] Speed ​​deviation is an instantaneous dynamic risk, the harm of which is immediately apparent when the limit is exceeded. The time window left for the system to react is extremely short. Therefore, the warning triggering conditions are relatively strict, allowing only 30% overshoot margin. Path deviation is a cumulative spatial risk, the harm of which intensifies as the deviation continues to increase. There is a certain buffer space during the deviation development process. Therefore, 50% overshoot margin is allowed to trigger a forced stop.

[0055] The trajectory deviation risk index is calculated by taking the maximum value among the speed deviation rate, path deviation rate, and positioning accuracy deviation rate. The overall safety status of trajectory motion is determined by the worst-case dimension; any deviation index exceeding the limit indicates a potential safety hazard. The comprehensive risk index should not be diluted by averaging high-risk dimensions simply because other dimensions perform well. A trajectory deviation risk index greater than or equal to 1 indicates that at least one dimension has an excess risk, while a value less than 1 indicates that all monitored dimensions are within the safe range.

[0056] Preferably, a motion trend prediction mechanism is introduced to predict the position one second in advance based on the velocity change trend within 0.5 seconds prior to the current moment. If the deviation between the predicted position and the template position exceeds 1.2 times the allowable deviation range of the path, an early warning is issued approximately 0.5 to 1 second before the deviation actually occurs.

[0057] The method further includes: S4: using depth point cloud to segment the patient's body surface point cloud, calculating the patient's overall center of gravity position; determining the bed support polygon, calculating the shortest distance from the center of gravity projection point to each side of the support polygon; defining the ratio of the shortest distance to the radius of the circumcircle of the bed base as the static stability margin; calculating the inertial force reduction, obtaining the dynamic stability margin, and outputting the overturning risk index.

[0058] In this embodiment, the depth point cloud data is preprocessed by using statistical filtering to remove outliers and voxel grid downsampling to reduce the point cloud density to a reasonable range. The preprocessed depth point cloud is then registered and identified using a pre-established 3D model of the operating table. Matched point clouds are assigned to the bed surface point cloud, while the remaining point clouds are assigned to the patient's body surface point cloud. Based on the patient's body surface point cloud, the overall center of gravity of the patient is calculated, and the coordinates of the center of gravity are transformed into the local coordinate system of the bed to obtain the coordinates of the center of gravity in the bed coordinate system. The arithmetic mean of the 3D coordinates of all points in the patient's body surface point cloud is then calculated to obtain the 3D coordinates of the center of gravity in the operating room world coordinate system. The local coordinate system of the bed is a coordinate system with the center of the bed surface as the origin, the major axis of the bed surface as the X-axis, the minor axis of the bed surface as the Y-axis, and the vertical axis upwards from the bed surface as the Z-axis.

[0059] In this embodiment, the bed support polygon is a quadrilateral formed by the contact points of the four casters of the operating bed base with the ground. The shortest distance from the center of gravity projection point (i.e., the projection of the center of gravity coordinates onto the XY plane of the bed's local coordinate system) to each side of the support polygon is calculated. The ratio of the shortest distance to the radius of the circumcircle of the bed base is defined as the static stability margin, with a value ranging from 0 to 1. This value characterizes the relative proximity of the current center of gravity projection point to the overturning boundary. A larger value indicates stronger anti-overturning capability under static conditions. When the ratio is 0, it indicates that the center of gravity projection has fallen on the support boundary and is in a critical overturning state.

[0060] Furthermore, dynamic corrections are introduced. Based on the output bed motion speed and acceleration, the influence of inertial force on stability is calculated to reflect the dynamic weakening effect of inertial force on stability during rapid body position adjustment.

[0061] When the bed undergoes lateral or longitudinal acceleration, the horizontal inertial force generated by the mass of the patient and the bed itself is calculated based on the bed's velocity and acceleration. ;in, It is a horizontal inertial force; The total mass of the patient and bed, in kilograms; It is the acceleration of the bed's motion, measured in meters per second².

[0062] The inertial force acts at the height of the center of gravity, generating a dynamic overturning moment relative to the support point (i.e., the overturning edge), which is: ;in, It is a horizontal inertial force; It is the dynamic overturning moment; It refers to the height of the center of gravity.

[0063] At the same time, gravity produces a static steady torque at the horizontal lever arm, which is: ;in, It is the stabilizing torque; g is the acceleration due to gravity; It is the horizontal lever arm, that is, the horizontal projection distance from the center of gravity to the overturning edge in the supporting polygon.

[0064] The reduction in stability margin due to inertial forces, i.e., the reduction in inertial forces, is defined as the ratio of the dynamic overturning moment to the static stabilizing moment: ;in, It is the reduction of inertial force; the greater the acceleration or the higher the center of gravity, the more significant the destructive effect of inertial force; the greater the horizontal lever arm (the center of gravity projection is far from the overturning boundary), the stronger the ability of the bed to resist dynamic disturbances.

[0065] The dynamic stability margin is the static stability margin minus the inertial force reduction. ;in, It is a dynamic stability margin; It is the static stability margin; It is the reduction in inertial force.

[0066] Under static conditions, medical electric beds typically require a static stability margin of at least 0.25 to meet clinical safety requirements. Considering sensor measurement errors, point cloud reconstruction accuracy errors, and operating room floor flatness deviations, the overall error budget is approximately ±5%. Simultaneously, to ensure sufficient stabilizing torque under dynamic disturbances to withstand unforeseen sudden external forces (such as accidental leaning by medical staff), at least 3% absolute stability margin is required as a physical safety baseline. Accordingly, a dynamic stability margin of 0.15 is set as the yellow warning threshold, corresponding to approximately 15% stability margin. At this point, inertial forces have consumed more than 40% of the static stability reserve, indicating a need for cautious operation. 0.08 is set as the orange warning threshold, corresponding to only 8% stability margin remaining. At this point, even minor disturbances may trigger instability, requiring immediate intervention. 0.03 is set as the red emergency threshold, where the stability margin is only 3%, below the overall error budget, indicating a physical critical state requiring unconditional emergency braking.

[0067] The system outputs a rollover risk index, calculated as 1 minus the dynamic stability margin, ranging from 0 to 1. A higher value indicates a higher rollover risk. A yellow warning is triggered when the rollover risk index exceeds 0.85, indicating caution is advised; an orange warning is triggered when the index exceeds 0.92, suggesting immediate cessation of positional adjustments; and a red warning is triggered when the index exceeds 0.97, automatically initiating emergency braking.

[0068] In conjunction with the initial enhanced image sequence, the method further includes: S5: estimating key points on the patient's body surface using human pose, calculating the slippage relative to the bed surface, and outputting a slippage risk index by combining a BMI adaptive threshold; slippage is sliding displacement.

[0069] A high-resolution network is employed as the backbone for human pose estimation. Parallel multi-resolution subnetworks maintain high-resolution feature representations, resulting in superior keypoint localization accuracy compared to traditional network architectures. After pre-training on a large human pose dataset, the network is fine-tuned using human images from an operating room setting to adapt to specific patient positions on the operating table.

[0070] Two-dimensional pixel coordinates of the following anatomical landmarks were extracted from multi-view RGB images: left acromion, right acromion, left anterior superior iliac spine, right anterior superior iliac spine, left knee joint, right knee joint, left heel, and right heel, totaling eight key points. For each key point, its three-dimensional coordinates in the operating room world coordinate system were reconstructed using triangulation with at least two viewpoints' two-dimensional coordinates and the corresponding camera's intrinsic and extrinsic parameter matrices.

[0071] By visually tracking the inherent features of the bed surface, including the seams between bed boards, the edge contours of the mattress, and circular positioning marks pre-drilled on the bed surface, a local coordinate system for the bed surface is established. The origin of the local coordinate system is set as the geometric center of the bed surface, with the X-axis along the major axis of the bed surface, the Y-axis along the minor axis of the bed surface, and the Z-axis perpendicular to the bed surface and upwards. The three-dimensional coordinates of key patient points are then transformed from the operating room world coordinate system to the local coordinate system of the bed surface.

[0072] Calculate the displacement vector of each key point relative to the initial moment of the patient's position (after the position is fixed and before the start of surgery). The core monitoring parameter for patient slippage risk is defined as the displacement of the shoulder along the long axis of the bed, in millimeters. In a head-down, feet-up position, the patient slips along the long axis of the bed due to the component of gravity; a continuous increase in displacement is the most direct indicator of slippage risk.

[0073] A baseline threshold of 20 mm was pre-set for slippage warning. Patient body shape adaptive adjustment was further introduced, calculating the Body Mass Index (BMI) based on the patient's preoperative height and weight. When the BMI reached or exceeded 30 (obese), the threshold was widened to 25 mm due to greater subcutaneous soft tissue slippage; when the BMI was below 18.5 (underweight), the threshold was tightened to 15 mm due to prominent skeletal landmarks and less soft tissue cushioning; for normal body types, the threshold remained at 20 mm.

[0074] A slip warning is triggered when shoulder displacement reaches or exceeds the adaptive threshold. Simultaneously, the slip velocity, i.e., the derivative of shoulder displacement with respect to time, is monitored. If the slip velocity consistently exceeds 0.5 mm per minute for more than 3 minutes, an early warning is triggered even if the displacement threshold has not yet been reached.

[0075] The system outputs a patient slippage risk index, calculated as the ratio of the current shoulder displacement to an adaptive threshold. A yellow alert is triggered when this ratio is greater than 0.7, and an orange alert is triggered when it reaches or exceeds 1.0.

[0076] S6: Based on steps S1 to S5, construct a multidimensional monitoring index set and generate an intraoperative monitoring and analysis report.

[0077] A multidimensional monitoring index set was constructed, including: bed posture data, micro-drift risk index, trajectory deviation risk index (only effective during active bed adjustments), overturning risk index, and patient slippage risk index.

[0078] If any index indicates a risk, a corresponding warning is issued directly; if all indices are within the normal range, a weighted composite score is calculated. The composite score is the sum of the products of each risk indicator and its corresponding weight, where the sum of the five weights is 1. The initial value of each weight is set to 0.2, and can be dynamically adjusted according to the type of surgery and the patient's condition for early warning and risk prevention.

[0079] After each surgery, a monitoring and analysis report is recorded and stored in the electronic health record of the operating table in a structured format. Each record includes a unique event number, timestamp, surgery type, patient basic information, warning level, risk attribution, intervention measures taken at the time, and subsequent results. The electric operating table is intelligently managed and controlled based on the monitoring data and warning status.

[0080] This application also includes constructing a bed-patient relative pose correlation model to distinguish between two fundamentally different phenomena: passive displacement of the patient caused by active bed movement and abnormal slippage of the patient when the bed is stationary. Specifically, it involves calculating the correlation coefficient between the change in bed pose and the change in displacement of key points on the patient.

[0081] The closer the absolute value of the correlation coefficient is to 1, the stronger the linear correlation between the two variables; the closer it is to 0, the weaker the linear correlation. A correlation threshold range is set, including an upper and lower threshold. When the correlation coefficient is greater than the upper threshold, it indicates a high correlation between patient displacement and bed motion, classifying it as passive displacement caused by active bed motion, a normal phenomenon that should not trigger a slippage warning. When the correlation coefficient is less than the lower threshold and the patient's key point displacement remains non-zero, it indicates that the patient is experiencing voluntary slippage while the bed is stationary, an abnormal phenomenon. The upper and lower thresholds are pre-set based on the significance level of the correlation coefficient in statistics and the engineering constraints of simultaneous multi-source sensor acquisition in the operating room environment. The square of the correlation coefficient represents the proportion of variance explained by the bed motion variable in the patient displacement variable. The coefficient of determination corresponding to the upper threshold indicates that bed motion can explain most of the variance in patient displacement, clinically manifested as a high degree of consistency in the movement trends of the two variables. The coefficient of determination corresponding to the lower threshold indicates that bed motion can only explain a very small portion of the variance in patient displacement, with the remainder attributed to non-bed-driven forces such as gravity slippage. In this embodiment, the upper threshold of correlation is preferably 0.7, and the lower threshold of correlation is preferably 0.3.

[0082] The transfer function between changes in the bed pitch angle and patient shoulder displacement is calculated to establish a spatial mapping relationship, i.e., the amount of patient displacement caused by a unit change in angle, expressed in millimeters per degree. The transfer function has different values ​​under different surgical positions; the value is larger in the head-down, feet-up position, and close to zero in the supine position.

[0083] The specific value of the transfer function is determined in real time as follows: Within a preset time window, the actual change in the bed's pitch angle and the corresponding measured displacement of the patient's shoulder are collected. Least squares linear regression is used to fit the proportional relationship between the two, and the slope coefficient obtained from the fitting is used as the estimated value of the transfer function under the current position. Based on the transfer function, a weighted calculation is performed on the current change in the bed's pitch angle to obtain the expected displacement of the patient's shoulder driven solely by bed movement.

[0084] The difference threshold is defined as the tolerance between the measured patient displacement and the expected displacement. When the absolute value of the difference exceeds the difference threshold, it indicates that there is an additional sliding driving force (such as gradual sliding caused by the component of gravity) in addition to the known driving force of bed movement. At this time, the associated abnormality warning is triggered.

[0085] The threshold for setting the difference is based on the following: under ideal rigid constraints, the relative displacement between the patient's surface landmarks and the bed surface should be zero; considering the elastic deformation of human soft tissue, the micro-movement of the skin relative to the bones, and the inherent error of the visual measurement system, a certain amount of deviation is allowed as a normal physiological-measurement composite tolerance. In this embodiment, the threshold for difference is preferably 5 mm. An associated abnormality index is output, calculated as the ratio of the absolute value of the difference between the measured patient displacement and the predicted displacement to the threshold for difference.

[0086] This embodiment achieves deep data fusion of bed status monitoring and patient status monitoring, filling the gap in existing solutions that only monitor the bed or patient independently without analyzing the correlation between the two, and providing key causal dimension information for comprehensive risk assessment of electric operating tables.

[0087] This application involves multiple spatial coordinate systems, including a pixel coordinate system, an image coordinate system, a camera coordinate system, an operating room world coordinate system, and a bed local coordinate system. The pixel coordinate system and the image coordinate system are used to describe the position of feature points in a two-dimensional image. The camera coordinate system uses the camera's optical center as its origin to describe the orientation of a point in three-dimensional space relative to the camera. The operating room world coordinate system uses the center of the operating room floor as its origin as a global reference benchmark. The bed local coordinate system is fixed to the operating table surface and moves with the bed, used to describe the patient's displacement relative to the bed surface.

[0088] The transformation between coordinate systems is achieved using a homogeneous transformation matrix, which is 4×4 in form and uniformly includes rotation and translation transformations, transforming 3D coordinate points in the source coordinate system to the target coordinate system. The camera intrinsic and extrinsic parameter matrices are pre-obtained using the Zhang calibration method, which is existing technology and will not be elaborated upon here. Typical transformation paths include: back-projecting pixel coordinates combined with depth values ​​to the operating room world coordinate system using the intrinsic and extrinsic parameter matrices; determining the pose of the origin and axis of the local bed coordinate system in the operating room world coordinate system by tracking bed surface feature points, and transforming the patient's key point coordinates from the operating room world coordinate system to the bed's local coordinate system; continuously transforming inertial measurement unit data to the operating room world coordinate system through gravity direction alignment during the initial static phase and the prediction equation of the extended Kalman filter. All sensor data are first transformed to the operating room world coordinate system before fusion calculation to ensure spatiotemporal reference consistency. Other related techniques can also be used for coordinate transformation, and this application does not impose specific limitations.

[0089] In this application, by acquiring operating table image sequences and preprocessing global illumination, stable enhanced image sequences can be obtained under complex lighting conditions such as strong light, shadow, and reflection in the operating room, providing high-quality image input. Through real-time target detection and 3D pose reconstruction of the operating table key point set, high-precision real-time estimation of the position and orientation of various parts of the table is achieved, providing a unified spatial benchmark for multi-dimensional monitoring and analysis.

[0090] By analyzing the time series of pitch and roll angles of the entire bed surface and backrest, and performing drift velocity and drift acceleration analysis, a linear extrapolation prediction model is constructed to achieve early identification of micro-drift trends and prediction of cumulative drift. Compared to existing technologies that can only detect obvious angular deviations that have already occurred, this invention can issue an early warning before the drift amount reaches the clinically dangerous threshold, allowing time for intervention.

[0091] By tracking the three-dimensional motion trajectory in real time during body position transitions and comparing the measured trajectory with a safe motion trajectory template frame by frame, a multi-dimensional quantitative assessment of speed deviation, path deviation, and positioning accuracy deviation is achieved. When the deviation exceeds the limit, a graded warning or motion pause is automatically triggered, effectively preventing patient injury caused by loss of movement control during body position adjustments.

[0092] By segmenting the deep point cloud to obtain the patient's surface point cloud, the overall center of gravity position of the patient is calculated and combined with the bed support polygon and inertial force correction to obtain the dynamic stability margin, thereby realizing the real-time quantitative assessment of the risk of tipping over during the body position adjustment process and overcoming the limitations of existing solutions that rely solely on static center of gravity judgment.

[0093] By estimating the three-dimensional coordinates of key points on the patient's body surface through human posture estimation and calculating the amount of slippage relative to the bed surface, combined with an adaptive BMI threshold, accurate monitoring of the risk of patient slippage along the bed surface is achieved, avoiding false alarms or missed alarms caused by differences in patient body size.

[0094] To make the technical solution of the present invention clearer, the above specific embodiments may involve several specific numerical values, thresholds, parameters, formulas, and examples. Those skilled in the art should understand that these specific contents are merely exemplary descriptions given for ease of implementation and are not essential technical features necessary for realizing the present invention, nor do they constitute a limitation on the scope of protection of the present invention. The mathematical formulas, parameter values, model structures, calculation rules, constraints, and data acquisition methods described in the specific embodiments section of this application are preferred examples for realizing the technical concept of the present invention, not the only implementation methods. Based on the core logic and inventive concept disclosed in this application, those skilled in the art can make reasonable equivalent substitutions, parameter adjustments, or logical optimizations to the parameters, formulas, and algorithm models according to actual conditions, industry standards, or conventional experimental methods, without producing substantial differences. Any changes that do not depart from the inventive concept of this application fall within the scope of this application. All mathematical expressions in this application should ensure that their physical meaning is clear when applied; parameters such as weights and thresholds can be calibrated using statistical or machine learning methods known in the art.

[0095] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A visual analysis-based intelligent monitoring method for an electric operating table, characterized in that, include: S1: Preprocess the global illumination of each operating table image to obtain an enhanced image sequence; Real-time target detection is performed on each image in the enhanced image sequence to identify the key point set of the operating table and obtain the pose vector of each key point; S2: Based on the pitch and roll angle time series of the entire bed surface and backboard in the pose vector, calculate the drift velocity and drift acceleration, predict the cumulative body position deviation at multiple key time points in the future, and output the micro-drift risk index. S3: During the body position transformation process, the pose vector is obtained and several local windows are determined. A three-dimensional motion trajectory is generated in each local window. The real-time three-dimensional motion trajectory is compared with the safe motion trajectory template frame by frame. The deviation index set is calculated and the trajectory deviation risk index is output. S4: Use depth point cloud to segment the patient's body surface point cloud and calculate the overall center of gravity position of the patient; Determine the supporting polygon of the bed, calculate the shortest distance from the center of gravity projection point to each side of the supporting polygon; define the ratio of the shortest distance to the radius of the circumcircle of the bed base as the static stability margin; calculate the inertial force reduction to obtain the dynamic stability margin, and output the overturning risk index. S5: Estimate key points on the patient's body surface using human posture, calculate slippage relative to the bed surface, and output slippage risk index by combining BMI adaptive threshold; S6: Based on steps S1 to S5, construct a multidimensional monitoring index set and generate an intraoperative monitoring and analysis report.

2. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 1, characterized in that, The method further includes: S51: Construct a bed-patient relative pose correlation model to distinguish between passive patient displacement caused by active bed movement and abnormal patient slippage when the bed is stationary; including: calculating the correlation coefficient between the change in bed pose and the change in patient key point displacement. A correlation threshold range is set, including an upper correlation threshold and a lower correlation threshold. When the correlation coefficient is greater than the upper correlation threshold, it is determined to be passive displacement caused by active bed movement, which is a normal phenomenon. When the correlation coefficient is less than the lower correlation threshold and the patient's key point displacement is continuously non-zero, it indicates that the patient has voluntary slippage while the bed is stationary, which is an abnormal phenomenon.

3. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 2, characterized in that, The method further includes: S52: Calculate the transfer function between the change in bed pitch angle and the patient's shoulder displacement, establishing a spatial mapping relationship, i.e., the amount of patient displacement caused by a unit angle change; the transfer function has different values ​​under different surgical positions. Based on the transfer function, the current change in bed pitch angle is weighted to obtain the expected displacement of the patient's shoulder driven only by bed movement; define the difference threshold as the deviation tolerance between the measured patient displacement and the expected displacement. When the absolute value of the difference exceeds the difference threshold, there is an additional slip driving force, triggering an association anomaly warning; output the association anomaly index, calculated as the ratio of the absolute value of the difference between the measured patient displacement and the predicted displacement to the difference threshold.

4. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 1, characterized in that, In step S1, the set of key points includes at least the headboard, backboard, seatboard, legboard, and the entire bed surface; For each key point, extract two-dimensional image feature points and back-project the two-dimensional image feature points onto the operating room world coordinate system to obtain the corresponding three-dimensional spatial coordinates; for each key point, define a six-degree-of-freedom pose vector, which includes the X, Y, and Z coordinates of the site center point in the operating room world coordinate system, as well as the roll angle, pitch angle, and yaw angle. The two-dimensional image feature points are the coordinates of the four corner points and the center point of the key point.

5. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 1, characterized in that, In step S2, the key time points include at least three time points with an interval of no less than three minutes; the cumulative body position offset includes the cumulative pitch angle offset and the cumulative roll angle offset; when the predicted cumulative offset at any key time point exceeds the preset safety boundary, it is determined that there is a micro-drift risk.

6. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 1, characterized in that, In step S3, the deviation index set includes speed deviation rate, path deviation rate, and positioning accuracy deviation rate; the trajectory deviation risk index is the maximum value among speed deviation rate, path deviation rate, and positioning accuracy deviation rate. The local window is the time period from the start time to the end time of the body position change. The start time is the moment when the position change rate of any part of the bed exceeds a preset speed threshold, and the end time is the moment when the position change rate of all parts is less than 0.1 degrees per second. Within the local window, the three-dimensional position coordinates and attitude angles of each part are recorded at fixed time intervals to form a three-dimensional motion trajectory.

7. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 1, characterized in that, In step S4, the bed support polygon is a quadrilateral formed by the contact points of the four casters of the operating bed base with the ground. Based on the bed's motion speed and acceleration, the horizontal inertial force generated by the mass of the patient and the bed itself is calculated. When the inertial force acts at the height of the center of gravity, it generates a dynamic overturning moment relative to the support point. At the same time, gravity generates a static stabilizing moment at the horizontal lever arm. The reduction in inertial force is defined as the ratio of the dynamic overturning moment to the static stabilizing moment. The dynamic stability margin is the static stability margin minus the reduction in inertial force. The overturning risk index is 1 minus the dynamic stability margin.

8. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 1, characterized in that, In step S5, the two-dimensional pixel coordinates of key points on the patient's body surface are extracted. The key points include the left acromion, right acromion, left anterior superior iliac spine, right anterior superior iliac spine, left knee joint, right knee joint, left heel, and right heel. For each key point, the three-dimensional coordinates in the operating room world coordinate system are reconstructed using the two-dimensional coordinates of at least two perspectives and the intrinsic and extrinsic parameter matrices of the corresponding camera through triangulation. Calculate the displacement vector of each key point relative to the initial moment of the body position, and define the monitoring parameter for the patient's slippage risk as the displacement of the shoulder along the long axis of the bed surface; When the shoulder displacement reaches the adaptive threshold, a slip warning is triggered; at the same time, the slip speed is monitored. Output the patient slippage risk index, calculated as the ratio of the current shoulder displacement to an adaptive threshold.

9. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 1, characterized in that, A pre-built safe motion trajectory template library is constructed, including: classifying standard body position transformation types according to the clinical use scenarios of electric operating tables, with each type having a clear combination of starting position, target position, and motion axis; under standard test conditions, the operator controls the operating table to complete the standard actions of each type of body position transformation, and the three-dimensional motion trajectory of each part is synchronously collected through the deployed visual perception system, with each type of action being repeated multiple times to cover the normal fluctuation range; multiple trajectories of the same type are spatiotemporally aligned; after alignment, the mean and standard deviation of the position coordinates at each sampling time are calculated, the mean position constitutes the ideal position time curve, and the mean position plus or minus twice the standard deviation constitutes the allowable deviation range of the path; The safe movement trajectory template includes: a body position transition type identifier, the ideal position-time curve of each body part during the transition process, the upper limit of speed, the upper limit of acceleration, and the allowable deviation range of the path.

10. The intelligent monitoring method for an electric operating table based on visual analysis according to claim 6, characterized in that, The speed deviation rate is the ratio of the difference between the real-time speed and the template speed to the upper speed limit; The path deviation rate is the ratio of the Euclidean distance between the real-time position and the ideal position of the template to the allowable path deviation under the current motion progress ratio; the positioning accuracy deviation rate is the ratio of the Euclidean distance between the final stopping position and the target position to the standard allowable value.