Operating room monitoring system based on real-time analysis of intraoperative operation data

Abstract

The invention relates to the technical field of data analysis, in particular to an operating room monitoring system based on intraoperative operation data real-time analysis, which comprises a real-time operation stage identification module, a dynamic weight configuration module, a multi-modal fusion analysis module and an intelligent early warning decision module, wherein the real-time operation stage recognition module generates a stage judgment signal by fusing an instrument movement track feature and an endoscope video stream feature; the dynamic weight configuration module activates a corresponding weight distribution strategy after receiving the stage judgment signal, and establishes a dynamic weight mapping relationship among the heterogeneous data; the multi-modal fusion analysis module performs time sequence alignment and weighted fusion on the heterogeneous data based on the dynamic weight mapping relationship, and outputs a multi-dimensional risk assessment vector; and the intelligent early warning decision module analyzes the multi-dimensional risk assessment vector and associates the context of the operation process, and triggers a differentiated alarm mode according to the risk grade gradient.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to an operating room monitoring system based on real-time analysis of intraoperative operation data. Background Art

[0002] Traditional operating room monitoring systems often rely on single-modality data (such as vital signs or imaging data) for independent analysis, making it difficult to capture the dynamic evolution of surgical scenarios in real time. With the widespread adoption of minimally invasive surgery and robotic-assisted technology, the complexity of intraoperative procedures has increased significantly. Existing systems have significant limitations in heterogeneous data integration, stage-adaptive monitoring, and risk warning. First, the spatiotemporal characteristics of multi-source data, such as instrument motion trajectories, endoscopic video streams, and physiological parameters, vary significantly, and the lack of an effective dynamic weighting mechanism makes it easy for key risk features to be overwhelmed by noise. Second, fixed-threshold alarm models struggle to adapt to the changing risk thresholds of different surgical stages, leading to false positives and missed alerts. Third, traditional single-dimensional risk assessments lack contextual analysis of the surgical process, making it impossible to establish a closed-loop feedback loop for the collaborative evolution of "operation-physiology-imaging."

[0003] In addition, clinical practice over-reliance on the experience and judgment of medical staff can easily lead to monitoring blind spots during long and complex operations. Summary of the Invention

[0004] The purpose of the present invention is to provide an operating room monitoring system based on real-time analysis of intraoperative operation data to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an operating room monitoring system based on real-time analysis of intraoperative operation data, comprising a real-time surgical stage identification module, a dynamic weight configuration module, a multimodal fusion analysis module, and an intelligent early warning decision module, wherein:

[0006] The real-time surgery stage recognition module generates a stage determination signal by fusing the instrument motion trajectory features and the endoscope video stream features;

[0007] The dynamic weight configuration module activates the corresponding weight allocation strategy after receiving the stage determination signal, and establishes a dynamic weight mapping relationship between heterogeneous data;

[0008] The multimodal fusion analysis module performs time series alignment and weighted fusion on heterogeneous data based on the dynamic weight mapping relationship, and outputs a multidimensional risk assessment vector;

[0009] The intelligent early warning decision module analyzes the multi-dimensional risk assessment vector and associates it with the surgical process context, triggering differentiated alarm modes according to the risk level gradient;

[0010] The system constructs a closed-loop feedback mechanism between the evolution of surgical scenarios and the adjustment of monitoring parameters through a cascade processing architecture that transforms stage identification into weighted mapping for fusion analysis and ultimately forms a decision response.

[0011] As a further improvement of this technical solution, the fusion process of the real-time surgical stage recognition module specifically includes:

[0012] Use position tracking technology to record the spatial movement path, speed, and motion pattern of surgical instruments to obtain the characteristics of instrument motion trajectory;

[0013] Apply image segmentation and object detection techniques to extract tissue structure change features from endoscopic image sequences and extract endoscopic video stream features;

[0014] The instrument motion trajectory features and the endoscope video stream features are fused into multimodal features to generate fused features.

[0015] As a further improvement of the present technical solution, the stage determination signal generation process includes: comprehensively analyzing the fusion features through a machine learning model, and outputting a stage determination signal containing a description of the surgical stage and prediction information.

[0016] As a further improvement of this technical solution, the process of acquiring the motion trajectory characteristics of the instrument specifically includes:

[0017] Use sensor equipment to collect equipment motion parameters in real time;

[0018] Establish the composite characteristic vector of device displacement-velocity-acceleration in a three-dimensional coordinate system;

[0019] The cutting and stitching operation features in the feature vector are extracted through pattern recognition algorithm.

[0020] As a further improvement of this technical solution, the endoscopic video stream feature extraction process specifically includes:

[0021] Use convolutional neural networks for real-time video frame feature extraction;

[0022] Mark the coordinates of bleeding points and area change rates through key point positioning technology;

[0023] Generate a time series feature matrix containing the degree of tissue deformation.

[0024] As a further improvement of the present technical solution, the heterogeneous data includes vital sign monitoring data, instrument mechanics data and imaging feature data.

[0025] As a further improvement of this technical solution, the weight allocation strategy process specifically includes:

[0026] Dynamically adjust the weight ratio of each data type according to the stage judgment signal;

[0027] Adaptive weight update mechanism based on real-time feature importance;

[0028] Weight correction feedback loop integrating expert knowledge base.

[0029] As a further improvement of the present technical solution, the intelligent early warning decision module includes an analysis unit, which is used to analyze the multidimensional risk assessment vector and associate it with the surgical process context, wherein the multidimensional risk assessment vector includes risk indicators of multiple dimensions, each dimension represents a risk factor, and each risk indicator has a corresponding numerical value representing its risk level.

[0030] As a further improvement of the present technical solution, the intelligent early warning decision module includes an alarm unit, which determines the risk level of each risk indicator based on the parsed risk assessment vector and associated context information. The risk levels include low risk, medium risk, and high risk.

[0031] As a further improvement of the present technical solution, the alarm unit triggers a differentiated alarm mode according to the risk level gradient, specifically including:

[0032] Low risk levels activate the visual prompt interface;

[0033] Medium risk level triggers an audible and visual composite alarm;

[0034] High risk levels trigger multimodal emergency alarms.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] This operating room monitoring system, which is based on real-time analysis of intraoperative operation data, is based on multimodal feature fusion and dynamic weight mapping, accurately identifying surgical stages and adaptively adjusting monitoring strategies, significantly improving the specificity of risk assessment in complex scenarios; time-aligned heterogeneous data fusion technology effectively captures cross-modal risk association characteristics and enhances early abnormality detection capabilities; the hierarchical warning mechanism is combined with surgical context analysis to achieve accurate responses from low-risk prompts to emergency interventions, and through stage identification, weight mapping to fusion analysis, and finally forming a cascade processing architecture for decision-making responses, a closed-loop feedback mechanism is constructed between the evolution of surgical scenes and the adjustment of monitoring parameters, which not only reduces the cognitive load of medical staff, but also continuously improves the level of intraoperative safety assurance through the system's self-optimization capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the overall module of the present invention;

[0038] Figure 2 This is a schematic diagram of the intelligent early warning decision module unit of the present invention.

[0039] In the figure: 100, real-time surgical stage recognition module; 200, dynamic weight configuration module; 300, multimodal fusion analysis module; 400, intelligent early warning decision module; 401, parsing unit; 402, alarm unit. DETAILED DESCRIPTION

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0041] Next, see Figure 1 The present invention provides a technical solution: an operating room monitoring system based on real-time analysis of intraoperative operation data, including a real-time surgical stage identification module 100, a dynamic weight configuration module 200, a multimodal fusion analysis module 300 and an intelligent early warning decision module 400.

[0042] The real-time surgical stage recognition module 100 generates a stage determination signal by fusing the instrument motion trajectory features and the endoscopic video stream features, specifically including:

[0043] Instrument motion trajectory characteristics refer to the movement paths, speeds, and motion patterns of various surgical instruments (such as scissors and clamps) used by doctors during surgery. High-precision position tracking technology or sensor equipment is used to record this information to form instrument motion trajectory characteristics. Analysis of the instrument motion trajectory characteristics can indirectly reflect the type or steps of the current operation. The specific implementation process includes:

[0044] Sensor equipment is used to collect instrument motion parameters in real time; a composite feature vector of instrument displacement, velocity and acceleration is established in a three-dimensional coordinate system; and cutting and suturing operation features are extracted from the feature vector using a pattern recognition algorithm.

[0045] Endoscopic video stream features refer to the continuous image sequence obtained from the endoscope, which can intuitively display the internal conditions of the surgical area. Processing methods include but are not limited to the application of computer vision technologies such as image segmentation, object detection, and key point positioning to extract useful information, such as changes in tissue structure and the location of bleeding points. It provides a window for directly observing the progress of the surgery and is very helpful for understanding the specific details of the operation. The specific implementation process includes:

[0046] Use convolutional neural networks for real-time video frame feature extraction;

[0047] Mark the coordinates of bleeding points and area change rates through key point positioning technology;

[0048] Generate a time series feature matrix containing the degree of tissue deformation.

[0049] Feature fusion and stage determination: First, the instrument trajectory features and endoscopic video stream features are preprocessed separately, such as removing noise and standardizing the format, to ensure the effectiveness of subsequent analysis. Then, key features are extracted from each type of data. For instrument trajectories, these are certain specific motion patterns; for video streams, the focus is on the state changes of specific organs or lesions.

[0050] The two sets of processed instrument motion trajectory features and endoscopic video stream features are combined to perform multimodal feature fusion to generate fused features. An appropriate machine learning model is used for comprehensive analysis. The goal here is to find an optimal combination so that the final output can most accurately reflect the current surgical stage.

[0051] Based on the above analysis results, the system will output a signal representing the current surgical stage as a stage determination signal. This stage determination signal not only contains a basic description of the surgical process status, but also contains some predictive information about possible events that may occur next, providing a basis for subsequent risk assessment.

[0052] After receiving the stage determination signal, the dynamic weight configuration module 200 activates the corresponding weight allocation strategy and establishes a dynamic weight mapping relationship between heterogeneous data. The specific process includes:

[0053] Receiving a stage determination signal: First, receiving a stage determination signal from the real-time surgery stage identification module 100 . This signal not only contains a basic description of the state of the surgery process, but also contains some prediction information about possible events that may occur next.

[0054] Obtain heterogeneous data based on the characteristics of the instrument's motion trajectory and the endoscope's video stream, including vital sign monitoring data, instrument mechanical data, and image feature data;

[0055] Activate the corresponding weight allocation strategy: Based on the received stage determination signal, the dynamic weight configuration module 200 selects and activates the weight allocation strategy that matches the current surgical stage. Different surgical stages may require attention to different types of data, so the weight allocation strategy is adjusted according to the different stages of the surgery. For example, in a specific surgical stage, vital sign monitoring data may be more important, while in another stage, instrument mechanical data or imaging feature data may be more critical. The weight allocation strategy dynamically adjusts the weight ratio of each data type based on the stage determination signal, and uses an adaptive weight update mechanism based on real-time feature importance, integrating the weight correction feedback loop of the expert knowledge base.

[0056] Preprocessing heterogeneous data: Before activating the weight allocation strategy, the vital sign monitoring data, instrument mechanical data, and imaging feature data in the heterogeneous data will be preprocessed, including noise removal and format standardization, to ensure data quality and consistency.

[0057] Extract key features: Extract key features for each type of data in heterogeneous data. For vital signs monitoring data, these are indicators such as heart rate and blood pressure; for instrument mechanical data, these are the force conditions and movement patterns of the instrument; for imaging feature data, these are changes in tissue structure and the location of bleeding points.

[0058] Establish dynamic weight mapping relationships: Combine processed heterogeneous data (vital sign monitoring data, instrument mechanical data, and imaging feature data) and use appropriate machine learning models for comprehensive analysis;

[0059] By analyzing the key features of heterogeneous data, the relative importance of each type of data in the current surgical stage is determined and a corresponding weight is assigned to it. This weight assignment is dynamic and can be automatically adjusted as the surgical stage changes; for example, in one stage, vital signs monitoring data may be given a higher weight, while in another stage, instrument mechanical data may be given a higher weight.

[0060] Output dynamic weight mapping relationship: Finally, a result representing the dynamic weight mapping relationship between various types of data is output. The result reflects which data are most critical in the current surgical stage and their respective weight ratios. The dynamic weight mapping relationship will serve as an important input for the subsequent multimodal fusion analysis module 300, and is used to perform time series alignment and weighted fusion of heterogeneous data, thereby generating a more accurate risk assessment vector.

[0061] The multimodal fusion analysis module 300 performs temporal alignment and weighted fusion on heterogeneous data based on a dynamic weight mapping relationship, and outputs a multidimensional risk assessment vector, specifically including:

[0062] Receive a dynamic weight mapping relationship from the dynamic weight configuration module 200, where the mapping relationship reflects the relative importance and weight ratio of vital sign monitoring data, instrument mechanical data, and imaging feature data in the current surgical stage;

[0063] Because different types of heterogeneous data (such as vital sign monitoring data, instrument mechanical data, and imaging feature data) have different sampling frequencies and timestamps, the multimodal fusion analysis module 300 needs to align these data on the time axis. Specific methods include interpolation, resampling, and other technologies to ensure the correspondence and synchronization of all data at the same time point for effective comprehensive analysis.

[0064] Appropriate machine learning models (such as deep neural networks, support vector machines, etc.) are used to conduct a comprehensive analysis of the weighted heterogeneous data. The goal of this model is to integrate these data and extract information that can fully reflect the surgical status and potential risks. Through this comprehensive analysis, various changes and abnormalities during the operation can be captured more accurately, thereby providing a more comprehensive risk assessment.

[0065] Based on the results of this comprehensive analysis, the multimodal fusion analysis module 300 generates a multidimensional risk assessment vector. This vector contains risk indicators across multiple dimensions, each corresponding to a specific risk factor. For example, the risk assessment vector might include dimensions such as heart rate abnormality risk, blood pressure fluctuation risk, instrument misuse risk, and bleeding risk. The risk indicator for each dimension reflects the severity of that risk factor at the current stage of surgery, providing key input for the subsequent intelligent early warning decision module 400.

[0066] See also Figure 2 The parsing unit 401 in the intelligent early warning decision module 400 parses the multi-dimensional risk assessment vector and associates it with the surgical process context, specifically including:

[0067] Receive a multi-dimensional risk assessment vector from the multimodal fusion analysis module 300, the vector including risk indicators of multiple dimensions, each dimension representing a specific risk factor, and each indicator having a corresponding numerical value indicating its risk level;

[0068] The multidimensional risk assessment vector is parsed to extract the specific values ​​of each risk indicator, which reflects the severity of various potential risks in the current surgical stage. For example, the risk assessment vector may include multiple dimensions such as abnormal heart rate risk, blood pressure fluctuation risk, improper instrument operation risk, and bleeding risk, and each dimension has a specific value.

[0069] The alarm unit 402 in the intelligent early warning decision module 400 triggers differentiated alarm modes according to the risk level gradient, specifically including:

[0070] The parsed risk assessment vector is associated with the contextual information of the current surgical stage, including but not limited to the current surgical stage, surgical type, patient medical history, and physician's operation notes. By associating this contextual information, the module can more accurately understand the specific meaning and impact of risk indicators. For example, an increase in a certain risk indicator at a specific surgical stage may have different meanings and treatment options.

[0071] Based on the parsed risk assessment vector and associated contextual information, the risk level of each risk indicator is determined. The risk level is usually divided into multiple gradients, including low risk, medium risk, and high risk.

[0072] The determination of risk levels is based on preset thresholds or dynamically adjusted thresholds. For example, the risk of abnormal heart rate may be set to a lower threshold in some surgical stages and a higher threshold in other stages.

[0073] According to the determined risk level gradient, the corresponding alarm mode is triggered. The alarm mode can be in various forms such as visual alarm, sound alarm, vibration alarm, etc., and differentiated processing can be carried out according to different risk levels;

[0074] The specific alarm modes are as follows:

[0075] Low risk: Activate the visual prompt interface to display a simple prompt message to remind medical staff to pay attention to changes in a certain indicator, but no obvious sound or light alarm will be issued;

[0076] Medium risk: triggers an audible and visual composite alarm to alert medical staff to further attention and treatment;

[0077] High risk: A multimodal emergency alarm is activated, with a strong audible alarm and a red alarm icon appearing on the display screen, possibly accompanied by vibration or other physical prompts to ensure that medical staff take immediate action.

[0078] Record the information of each alarm, including alarm time, risk type, risk level, treatment measures, etc. These records can be used for subsequent analysis and improvement; at the same time, they can also provide real-time feedback to help medical staff understand the current treatment effect and adjust the alarm strategy according to actual conditions.

[0079] Through the above process, the intelligent early warning decision module 400 can effectively analyze the multi-dimensional risk assessment vector, and combine the context information of the surgical process to trigger differentiated alarm modes according to the risk level gradient, thereby improving the safety and efficiency of the operation. This not only helps to timely discover and deal with potential risks, but also improves the controllability and safety of the entire surgical process.

[0080] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely preferred examples of the present invention and are not intended to limit the present invention. Various changes and improvements may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and improvements fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. An operating room monitoring system based on real-time analysis of intraoperative operation data, characterized by: The system comprises a real-time surgical stage identification module (100), a dynamic weight configuration module (200), a multimodal fusion analysis module (300) and an intelligent early warning decision module (400), wherein: The real-time surgical stage recognition module (100) generates a stage determination signal by fusing instrument motion trajectory features and endoscope video stream features; The dynamic weight configuration module (200) activates the corresponding weight allocation strategy after receiving the stage determination signal, and establishes a dynamic weight mapping relationship between heterogeneous data; The multimodal fusion analysis module (300) performs time sequence alignment and weighted fusion on heterogeneous data based on the dynamic weight mapping relationship, and outputs a multidimensional risk assessment vector; The intelligent early warning decision module (400) analyzes the multi-dimensional risk assessment vector and associates it with the surgical process context, triggering a differentiated alarm mode according to the risk level gradient; The system constructs a closed-loop feedback mechanism between the evolution of surgical scenarios and the adjustment of monitoring parameters through a cascade processing architecture that transforms stage identification into weighted mapping for fusion analysis and ultimately forms a decision response.

2. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 1, characterized in that: The fusion process of the real-time surgical stage identification module (100) specifically includes: Use position tracking technology to record the spatial movement path, speed, and motion pattern of surgical instruments to obtain the characteristics of instrument motion trajectory; Apply image segmentation and object detection techniques to extract tissue structure change features from endoscopic image sequences and extract endoscopic video stream features; The instrument motion trajectory features and the endoscope video stream features are fused into multimodal features to generate fused features.

3. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 1, characterized in that: The stage determination signal generation process includes: comprehensively analyzing fusion features through a machine learning model, and outputting a stage determination signal containing a description of the surgical stage and prediction information.

4. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 3 is characterized in that: The process of acquiring the characteristics of the motion trajectory of the instrument specifically includes: Use sensor equipment to collect equipment motion parameters in real time; Establish the composite characteristic vector of device displacement-velocity-acceleration in a three-dimensional coordinate system; The cutting and stitching operation features in the feature vector are extracted through pattern recognition algorithm.

5. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 3 is characterized in that: The endoscope video stream feature extraction process specifically includes: Use convolutional neural networks for real-time video frame feature extraction; Mark the coordinates of bleeding points and area change rates through key point positioning technology; Generate a time series feature matrix containing the degree of tissue deformation.

6. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 1, characterized in that: The heterogeneous data includes vital sign monitoring data, instrument mechanics data and imaging feature data.

7. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 1, characterized in that: The weight allocation strategy process specifically includes: Dynamically adjust the weight ratio of each data type according to the stage judgment signal; Adaptive weight update mechanism based on real-time feature importance; Weight correction feedback loop integrating expert knowledge base.

8. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 1, characterized in that: The intelligent early warning decision module (400) includes a parsing unit (401), which is used to parse a multidimensional risk assessment vector and associate it with a surgical process context, wherein the multidimensional risk assessment vector includes risk indicators of multiple dimensions, each dimension represents a risk factor, and each risk indicator has a corresponding numerical value representing its risk level.

9. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 1, characterized in that: The intelligent early warning decision module (400) includes an alarm unit (402), and the alarm unit (402) determines the risk level of each risk indicator based on the parsed risk assessment vector and associated context information, where the risk level includes low risk, medium risk, and high risk.

10. The operating room monitoring system based on real-time analysis of intraoperative operation data according to claim 9, characterized in that: The alarm unit (402) triggers a differentiated alarm mode according to the risk level gradient, specifically including: Low risk levels activate the visual prompt interface; Medium risk level triggers an audible and visual composite alarm; High risk levels trigger multimodal emergency alarms.

Images