CN-122022193-A - Microscopic endoscope operation evaluation method and system

CN122022193ACN 122022193 ACN122022193 ACN 122022193ACN-122022193-A

Abstract

The invention provides a microscopic endoscope operation evaluation method and system, comprising the steps of preprocessing and extracting features of a microscopic endoscope operation data stream, generating an original motion primitive sequence and a device event sequence, acquiring comprehensive objective performance scores in an operation analysis window according to preset microscopic endoscope operation primitive indexes, synchronously combining a dynamic efficient operation library, generating conditional pareto fronts of various anatomical parts, acquiring environmental state data, analyzing operation to be evaluated by combining a preset strategy performance evaluation mechanism, acquiring a strategy performance evaluation result, performing causal analysis on the operation to be evaluated based on a Bayesian structural equation model, performing inverse fact attribution by combining a dynamic efficient operation library, generating a causal analysis data set, performing comprehensive operation evaluation according to the comprehensive objective performance scores, the conditional pareto fronts, the strategy performance evaluation result and the causal analysis data set, and generating a comprehensive operation evaluation result, thereby improving accuracy of microscopic endoscope operation evaluation.

Inventors

MAO YING
ZHUANG QIYUAN
CHEN LIANG
ZOU XIANG

Assignees

复旦大学附属华山医院

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. A method for evaluating operation of a micro-endoscope, comprising the steps of: Preprocessing the collected microscopic endoscope operation data stream, and extracting features of the preprocessed data stream to generate an original motion primitive sequence and a device event sequence; According to a preset microscopic endoscope operation element index, acquiring comprehensive objective performance scores in an operation analysis window, and synchronously combining with a dynamic high-efficiency operation library to generate a condition pareto front edge of each anatomical part; Acquiring environmental state data corresponding to a microscopic endoscope operation data stream, and analyzing the operation to be evaluated by combining a preset strategy efficiency evaluation mechanism to acquire a strategy efficiency evaluation result; Performing causal analysis on the operation to be evaluated based on a Bayesian structural equation model, and performing inverse fact attribution by combining the dynamic efficient operation library to generate a causal analysis data set; And carrying out comprehensive operation evaluation according to the comprehensive objective performance score, the conditional pareto leading edge, the strategy performance evaluation result and the causal analysis data set to generate a comprehensive operation evaluation result.
2. The method of claim 1, wherein preprocessing the acquired microscopic endoscope operation data stream and extracting features of the preprocessed data stream to generate the original motion primitive sequence and the device event sequence, and comprising: The micro-endoscope operation data stream at least comprises an original motion signal sequence, an equipment operation log and a corresponding time stamp; dividing the original motion signal sequence into a plurality of motion segments with fixed sizes, and extracting the characteristics of each motion segment to obtain a motion segment characteristic set; Clustering the motion segment feature set by adopting a Gaussian mixture model to generate a plurality of original motion pattern clusters; Processing the original motion signal sequence based on the original motion pattern cluster to generate an original motion primitive sequence; And extracting the equipment events which are time-aligned with the original motion primitive sequence from the equipment operation log, and generating an equipment event sequence.
3. The method for evaluating operation of a micro-endoscope according to claim 1, wherein the steps of obtaining comprehensive objective performance scores in an operation analysis window according to a preset micro-endoscope operation element index, synchronously combining with a dynamic high-efficiency operation library, and generating a conditional pareto front of each anatomical part comprise the steps of: the preset microscopic endoscope operation element index at least comprises an information integrity index, target capturing efficiency and a physiological disturbance index; analyzing the current microscopic endoscope operation video stream by adopting a mode of combining visual SLAM with image definition to acquire an information integrity index; combining a pre-trained focus detection model, obtaining the focus proportion effectively captured by an operator in the current microscopic endoscope operation, and generating target capturing efficiency based on the effectively captured focus proportion; acquiring a physiological disturbance index of the current microscopic endoscope operation by adopting optical flow and tissue deformation analysis; dividing the current operation of the micro endoscope into a plurality of operation analysis windows according to the preset operation analysis window capacity; acquiring comprehensive objective performance scores based on preset microscopic endoscope operation element indexes in the operation analysis window; And combining the dynamic high-efficiency operation library to generate a conditional pareto front edge for the corresponding anatomical part of the current microscopic endoscope operation.
4. A method of evaluating a microscopic endoscopic procedure according to claim 3, wherein generating a conditional pareto front for a current microscopic endoscopic procedure corresponding to an anatomical site in combination with a dynamic efficient procedure library comprises: The dynamic efficient operation library is expressed as an efficient operation database constructed based on expert demonstration operation data, and is updated periodically based on the comprehensive objective performance score; the dynamic high-efficiency operation library is used for storing high-efficiency microscopic endoscope operations, and each high-efficiency microscopic endoscope operation at least comprises a corresponding original operation fragment, a comprehensive objective efficacy score and a description label; In the dynamic efficient operation library, grouping data in the library based on preset anatomical part labels to generate a plurality of anatomical part groups; And for the anatomical part group, acquiring the pareto front of the corresponding condition based on a preset microscopic endoscope operation element index in an operation analysis window.
5. The method for evaluating microscopic endoscope operation according to claim 1, wherein obtaining environmental state data corresponding to a microscopic endoscope operation data stream, analyzing an operation to be evaluated in combination with a preset policy efficacy evaluation mechanism, obtaining a policy efficacy evaluation result, comprises: Extracting environmental state characteristics of a video frame corresponding to the micro-endoscope operation data stream to obtain environmental state data; The environmental state data at least comprises exploration completeness and local task complexity; Training an operator policy network based on the microscopic endoscope operation data stream, the operator policy network for simulating a decision mechanism of an operator; And carrying out policy performance evaluation and policy environment stability analysis based on the operator policy network to obtain a policy performance evaluation result.
6. The method of claim 5, wherein performing a policy performance evaluation and a policy environmental stability analysis based on the operator policy network, obtaining a policy performance evaluation result, comprises: performing policy efficacy evaluation based on a preset policy efficacy evaluation index set to obtain a policy efficacy index; the preset strategy efficiency evaluation index set at least comprises strategy entropy, environment action mutual information and target guidance index; the operator strategy network carries out simulation prediction based on the environmental state data to generate a simulation action sequence; Injecting disturbance into the environmental state data to generate disturbance environmental state data; the operator strategy network carries out simulation prediction based on the disturbance environment state data to generate a disturbance simulation action sequence; analyzing the simulation action sequence and the disturbance simulation action sequence based on a preset strategy environment stability index set to obtain a strategy stability index; and generating a strategy efficiency evaluation result based on the strategy efficiency index and the strategy stability index.
7. The method of claim 1, wherein performing causal analysis on the operation to be evaluated based on a bayesian structural equation model and performing anti-facts attribution in combination with the dynamic efficient operation library, generating a causal analysis dataset comprises: Carrying out intention reasoning on the original motion primitive sequence based on a hidden Markov model to acquire an intention variable; Extracting action variables and response variables from a microscopic endoscope video frame sequence corresponding to the operation to be evaluated; the action variable is used for representing an action executed by an operator; the response variable is used for representing the quality of a microscopic endoscope video frame; obtaining a perception variable according to a preset perception analysis mechanism, wherein the perception variable is used for representing the perception state of an operator; Constructing a Bayesian structure equation model based on the intention variable, the extracted action variable, the response variable and the perception variable; estimating the Bayesian structure equation model by adopting a Markov chain Monte Carlo method to obtain a causal estimation parameter set.
8. A method for evaluating a micro-endoscopic procedure according to claim 7, characterized in that the method further comprises: Performing low-efficiency period identification on the current microscopic endoscope operation process based on the comprehensive objective efficiency score to obtain a low-efficiency period set; For each low-efficiency period in the low-efficiency period set, selecting an ideal reference segment from the dynamic high-efficiency operation library; performing inverse fact simulation in a corresponding low-efficiency period based on the ideal reference segment, and obtaining an inverse fact attribution result; the causal estimate parameter set and the anti-facts attribution result are packaged as a causal analysis data set.
9. The method of claim 1, wherein performing a comprehensive operation assessment based on the comprehensive objective performance score, the conditional pareto front, the strategic performance assessment result, and the causal analysis dataset to generate a comprehensive operation assessment result, comprising: Acquiring an operation efficiency score based on the conditional pareto front; obtaining a strategy efficiency score according to the strategy efficiency evaluation result; generating a causal link score based on the causal analysis dataset; Generating a comprehensive operation score and an operation improvement suggestion based on the comprehensive objective performance score, the operation efficiency score, the strategy performance score and the causal relationship score; And packaging the comprehensive operation score and the operation improvement suggestion into a comprehensive operation evaluation result.
10. A micro-endoscope operation evaluation system for implementing the micro-endoscope operation evaluation method according to any one of claims 1 to 9, comprising: the feature extraction module is used for preprocessing and extracting features of the micro-endoscope operation data stream to obtain the original motion primitive sequence and the equipment event sequence; The operation evaluation module is used for acquiring the comprehensive objective performance scores in an operation analysis window and generating the condition pareto front edge of each anatomical part by combining a dynamic high-efficiency operation library; The strategy evaluation module is used for acquiring environmental state data, analyzing the operation to be evaluated by combining a preset strategy efficiency evaluation mechanism and acquiring the strategy efficiency evaluation result; the cause and effect analysis module is used for carrying out cause and effect analysis on the operation to be evaluated, carrying out inverse fact attribution by combining the dynamic high-efficiency operation library and generating the cause and effect analysis data set; The comprehensive evaluation module is used for carrying out comprehensive operation evaluation according to the comprehensive objective efficacy score, the conditional pareto front edge, the strategy efficacy evaluation result and the causal analysis data set to generate the comprehensive operation evaluation result.

Description

Microscopic endoscope operation evaluation method and system Technical Field The invention relates to the technical field of image analysis, in particular to a microscopic endoscope operation evaluation method and system. Background Quantitative assessment and effective training of microscopic endoscope operation skills are important components in clinical medical education, however, the current traditional assessment method has certain limitations. In particular, traditional assessment methods mostly stay at the descriptive statistical level, often neglecting modeling of the policy logic contained in the operator behavior, and at the same time, often lacking modeling of interpretable causal links from raw data to clinical errors, resulting in insufficient accuracy in assessment of microscopic endoscopic operations. Disclosure of Invention Aiming at the defects existing in the prior art, the embodiment of the invention provides a microscopic endoscope operation evaluation method and a microscopic endoscope operation evaluation system, which are used for solving at least one of the technical problems existing in the prior art. In order to solve the technical problems, the technical scheme adopted by the invention is as follows: In a first aspect, the present invention provides a method of evaluating a microscopic endoscopic procedure, comprising: Preprocessing the collected microscopic endoscope operation data stream, and extracting features of the preprocessed data stream to generate an original motion primitive sequence and a device event sequence; According to a preset microscopic endoscope operation element index, acquiring comprehensive objective performance scores in an operation analysis window, and synchronously combining with a dynamic high-efficiency operation library to generate a condition pareto front edge of each anatomical part; Acquiring environmental state data corresponding to a microscopic endoscope operation data stream, and analyzing the operation to be evaluated by combining a preset strategy efficiency evaluation mechanism to acquire a strategy efficiency evaluation result; Performing causal analysis on the operation to be evaluated based on a Bayesian structural equation model, and performing inverse fact attribution by combining the dynamic efficient operation library to generate a causal analysis data set; And carrying out comprehensive operation evaluation according to the comprehensive objective performance score, the conditional pareto leading edge, the strategy performance evaluation result and the causal analysis data set to generate a comprehensive operation evaluation result. In a possible implementation manner of the first aspect, preprocessing an acquired microscopic endoscope operation data stream, and extracting features of the preprocessed data stream to generate an original motion primitive sequence and a device event sequence, including: The micro-endoscope operation data stream at least comprises an original motion signal sequence, an equipment operation log and a corresponding time stamp; dividing the original motion signal sequence into a plurality of motion segments with fixed sizes, and extracting the characteristics of each motion segment to obtain a motion segment characteristic set; Clustering the motion segment feature set by adopting a Gaussian mixture model to generate a plurality of original motion pattern clusters; Processing the original motion signal sequence based on the original motion pattern cluster to generate an original motion primitive sequence; And extracting the equipment events which are time-aligned with the original motion primitive sequence from the equipment operation log, and generating an equipment event sequence. In a possible implementation manner of the first aspect, according to a preset micro-endoscope operation element index, a comprehensive objective performance score in an operation analysis window is obtained, and a dynamic efficient operation library is synchronously combined to generate a conditional pareto front edge of each anatomical part, including: the preset microscopic endoscope operation element index at least comprises an information integrity index, target capturing efficiency and a physiological disturbance index; analyzing the current microscopic endoscope operation video stream by adopting a mode of combining visual SLAM with image definition to acquire an information integrity index; combining a pre-trained focus detection model, obtaining the focus proportion effectively captured by an operator in the current microscopic endoscope operation, and generating target capturing efficiency based on the effectively captured focus proportion; acquiring a physiological disturbance index of the current microscopic endoscope operation by adopting optical flow and tissue deformation analysis; dividing the current operation of the micro endoscope into a plurality of operation analysis windows according to the preset operatio