Hepatobiliary artery suture hemostasis operation training result discrimination method and training model

Abstract

The invention relates to a hepatobiliary artery suture hemostasis operation training result discrimination method and a training model. The method comprises the following steps: firstly, acquiring operation training data related to physiological information of tissues, operation data of surgical instruments and image information of an operation area, and then extracting the acquired training data by using a data processing technology so as to extract feature information closely related to operation training quality; and finally, inputting the feature information of the operation training result to be discriminated into a discrimination model trained by a plurality of machine learning classifiers, and outputting an operation training result discrimination result including grade evaluation for judging whether the operation is qualified or not by the model, thereby realizing scientific and accurate evaluation and judgment on the operation training effect.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence, and particularly relates to a method for discriminating the training results of hepato-biliary artery suture hemostasis surgery and a training model. Background Art

[0002] With the development of artificial intelligence technology, there has emerged a technology for discriminating the training results of hepato-biliary artery suture hemostasis surgery. The operation of hepato-biliary artery suture hemostasis surgery is extremely difficult and risky, and its success or failure is directly related to the life health and prognosis of patients. Traditional surgical training methods, such as relying on limited cadaver specimens for training, are restricted by factors such as difficult acquisition of specimens and large differences from living bodies; animal experiment training cannot accurately simulate the human surgical situation due to different anatomical structures and animal welfare issues; clinical teaching cannot give novices sufficient practice opportunities due to surgical risks. At the same time, existing training evaluations are mostly based on subjective experience, with lagging real-time feedback and low intelligence level, lacking objective quantitative criteria, and it is difficult to achieve scientific and accurate evaluation and judgment of surgical training effects. Summary of the Invention

[0003] Based on this, in view of the above technical problems, it is necessary to provide a method for discriminating the training results of hepato-biliary artery suture hemostasis surgery and a training model that can achieve scientific and accurate evaluation and judgment of surgical training effects.

[0004] In a first aspect, the present application provides a method for discriminating the training results of hepato-biliary artery suture hemostasis surgery, including:

[0005] Obtaining surgical training data related to hepato-biliary artery suture hemostasis surgery; the surgical training data includes the physiological information of tissues, the operation data of surgical instruments, and the image information of the surgical operation area.

[0006] Processing the collected surgical training data to obtain feature information related to the quality of surgical training.

[0007] Inputting the feature information of the surgical training result to be discriminated into the trained discrimination model to obtain the discrimination result of the surgical training result; the discrimination result includes a grade evaluation for judging whether the surgical operation is qualified; wherein, the discrimination model is trained by combining multiple machine learning classifiers.

[0008] In one embodiment, processing the collected surgical training data to obtain feature information related to the quality of surgical training includes:

[0009] Performing noise reduction and filtering processing on the physiological information of tissues and the operation data of surgical instruments to obtain processed physiological parameters and operation parameters.

[0010] Perform image enhancement processing, denoising, and grayscale processing on the image information of the surgical operation area to obtain the processed image parameters.

[0011] Construct a data set based on each data parameter in the physiological parameters, operation parameters, and image parameters to obtain the sample data to be processed.

[0012] Perform normalization and outlier removal operations on the sample data to be processed to obtain a feature data set.

[0013] Balance the data based on the feature data set using the Synthetic Minority Over-sampling Technique (SMOTE) to obtain a balanced data set.

[0014] Use the trained deep learning model to extract features from the balanced data set to obtain the corresponding feature information; the feature information includes the rationality feature of tissue processing, the hemostasis effect feature, the stability feature for evaluating the operation stability of surgical instruments, and the accuracy feature of suture needle puncture and suture.

[0015] Use the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm to assist in verifying the feature information to obtain updated feature information.

[0016] In one embodiment, use the trained deep learning model to extract features from the balanced data set to obtain the corresponding feature information, including:

[0017] Process the data in the balanced data set using the Recursive Feature Elimination (RFE) method to obtain different feature subsets.

[0018] Use the Multi-Layer Perceptron (MLP) algorithm to integrate different feature subsets to obtain a comprehensive feature set that combines all data parameters.

[0019] Perform k-fold cross-validation on the integrated comprehensive feature set to obtain the corresponding performance metrics; the performance metrics include the rationality metric of tissue processing, the hemostasis effect metric, the stability metric for evaluating the operation stability of surgical instruments, and the accuracy metric of suture needle puncture and suture.

[0020] Calculate the performance metrics to evaluate the contribution degree of each performance metric to the discriminant model to obtain the feature evaluation value.

[0021] Screen the feature evaluation values to obtain the feature information that meets the conditions of the preset evaluation metrics.

[0022] In one embodiment, calculate the performance metrics to evaluate the contribution degree of each performance metric to the discriminant model to obtain the feature evaluation value, including:

[0023] Calculate based on the rationality metric of tissue processing using a formula to obtain the corresponding rationality feature evaluation value.

[0024]

[0025] Among them, S t represents the rationality characteristic evaluation value, k represents the number of monitoring points, l represents the actual tissue deformation degree of the i-th monitoring point, T n represents the normally allowed tissue deformation degree, F l represents the actual tissue force of the i-th monitoring point, F n represents the safe force range that the tissue can withstand, V l represents the tissue blood flow velocity change rate of the i-th monitoring point, V n represents the acceptable blood flow velocity change rate range.

[0026] Based on the hemostasis effect index of tissue treatment, it is calculated using a formula to obtain the corresponding hemostasis effect characteristic evaluation value.

[0027]

[0028] Among them, S h represents the hemostasis effect characteristic evaluation value, t represents the bleeding stop time, t s represents the preset standard hemostasis time, A b represents the final bleeding area, A a represents the maximum acceptable bleeding area.

[0029] Based on the stability index for evaluating the operation stability of surgical instruments, it is calculated using a formula to obtain the corresponding stability characteristic evaluation value.

[0030]

[0031] Among them, S a represents the surgical instrument operation stability score, n represents the number of sampling points, F i , D i , A i respectively represent the force, displacement, and angle values of the i-th sampling point, represents the average value of force, displacement, and angle, F max , D max , A max represent the maximum values of force, displacement, and angle.

[0032] Based on the accuracy index of suture needle puncture and suture, it is calculated using a formula to obtain the corresponding accuracy characteristic evaluation value.

[0033]

[0034] Among them, S c represents the suture needle puncture accuracy characteristic evaluation value, m represents the number of punctures, P jrepresents the deviation distance between the actual position and the target position of the \(i\)-th puncture, \(P\) t represents the preset maximum acceptable deviation distance, \(B\) j represents the actual bending angle of the suture needle during the \(i\)-th puncture, \(B\) t represents the preset ideal bending angle range.

[0035] In one embodiment, the feature information for judging the surgical training result is input into the trained discrimination model to obtain the discrimination information of the surgical training result, including:

[0036] Data annotation is performed on each feature information according to the fuzzy comprehensive evaluation method to obtain the feature category parameters corresponding to each feature information.

[0037] Missing value processing is performed on the processed feature category parameters to obtain the missing value analysis result.

[0038] Among them, when the missing value ratio of all feature category parameters is less than 10%, the missing value analysis result is abnormal. For the missing feature category parameters, the mode filling method is used for data filling. For continuous feature category parameters, the random forest algorithm is used for imputation.

[0039] When the missing value ratio of all feature category parameters is greater than or equal to 10%, the missing value analysis result is normal, and the feature category parameters are kept unchanged.

[0040] Based on the missing value analysis result, the feature category parameters are updated to obtain the updated feature category parameters.

[0041] Each updated feature category parameter is input into the trained discrimination model to obtain the discrimination information of the surgical training result.

[0042] Based on the grade evaluation formula, the discrimination information is classified by grade to obtain the discrimination result for classifying and discriminating the surgical training result.

[0043] In one embodiment, the trained discrimination model is obtained through the following steps, including:

[0044] Use a machine learning classifier for model training to obtain a discrimination classifier; among them, the machine learning classifier is at least one of Gaussian Naive Bayes, neural network, ridge regression, and linear logistic regression.

[0045] Use the grid search method to train each discrimination classifier to obtain a pre-trained discrimination model.

[0046] Based on the model trained with historical data, the feature information is normalized to obtain a feature typing sample set.

[0047] Process each feature classification sample set until the feature evaluation values corresponding to each feature information meet the preset evaluation index conditions, and obtain the balanced feature classification sample set.

[0048] Use each feature information in the balanced feature classification sample set as the input of the current pre-trained discriminant model with determined initial values, and use the discriminant information corresponding to each feature information as the expected output of the current pre-trained discriminant model with determined initial values.

[0049] Construct a loss function for the actual output and the expected output of the current discriminant model, perform backpropagation based on the loss function, and modify the undetermined network parameters of the current pre-trained discriminant model until the loss function meets the set requirements, and complete the training of the current pre-trained discriminant model.

[0050] In one embodiment, perform grade classification on the discriminant information based on the grade evaluation formula to obtain the discriminant result for classifying and discriminating the surgical training result, including:

[0051] Use the following formula to calculate the discriminant information to evaluate the influence of multiple key factors on the surgical effect, and obtain the comprehensive score of the surgical training result.

[0052]

[0053] Among them, S represents the comprehensive score of the surgical training result, n represents the number of feature category parameters affecting the surgical training result, w i represents the weight of the i-th feature category parameter, S i represents the feature evaluation value of the i-th feature category parameter, f i (S i ) represents the score conversion function for the i-th feature category parameter, which converts the feature evaluation value into a score within the range of [0, 1].

[0054] Judge the comprehensive score of the surgical training result based on the preset scoring interval to obtain the discriminant result of the classification discrimination.

[0055] In the second aspect, the present application also provides a training model for discriminating the surgical training result of hepatobiliary artery suture hemostasis, and the training model includes:

[0056] A data acquisition module for acquiring surgical training data related to hepatobiliary artery suture hemostasis surgical training; the surgical training data includes the physiological information of tissues, the operation data of surgical instruments, and the image information of the surgical operation area.

[0057] A data processing module for processing the collected surgical training data to obtain feature information related to the surgical training quality.

[0058] A model training module for inputting the feature information of the surgical training result to be discriminated into a trained discrimination model to obtain the discrimination result of the surgical training result; the discrimination result includes a grade evaluation for judging whether the surgical operation is qualified; wherein, the discrimination model is trained by combining multiple machine learning classifiers.

[0059] In a third aspect, the present application also provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the method as described above are implemented.

[0060] In a fourth aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method as described above are implemented.

[0061] For the above-mentioned method for discriminating the surgical training result of hepato-biliary artery suture hemostasis, training model, computer device and storage medium, first, surgical training data including the physiological information of tissues, the operation data of surgical instruments and the image information of the surgical operation area is obtained, and then data processing technology is used to extract the collected training data, so as to refine the feature information closely related to the surgical training quality. Finally, the feature information of the surgical training result to be discriminated is input into a discrimination model trained by multiple machine learning classifiers, and the discrimination result of the surgical training result including the grade evaluation for judging whether the surgical operation is qualified is output by the model, so as to realize the scientific and accurate evaluation and judgment of the surgical training effect. Description of the Drawings

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0063] Figure 1 It is a flowchart of a method for discriminating the surgical training result of hepato-biliary artery suture hemostasis provided by an embodiment of the present invention;

[0064] Figure 2 It is a flowchart of processing the collected surgical training data to obtain the feature information related to the surgical training quality provided by an embodiment of the present invention;

[0065] Figure 3 It is a flowchart of using a trained deep learning model to extract features from a balanced data set to obtain the corresponding feature information provided by an embodiment of the present invention;

[0066] Figure 4The flowchart of the step of inputting the feature information of the surgical training result to be discriminated into the trained discrimination model to obtain the discrimination information of the surgical training result provided by the embodiment of the present invention;

[0067] Figure 5 The flowchart of obtaining the trained discrimination model provided by the embodiment of the present invention;

[0068] Figure 6 The structural block diagram of a training model for discriminating the surgical training result of hepato-biliary artery suture hemostasis provided by the embodiment of the present invention. Detailed implementation manners

[0069] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0070] Secondly, the implementation environment of the embodiment of the present application will be described. Exemplarily, the implementation environment includes a data acquisition hardware 101, a data processing device 102, and an intelligent device terminal 103.

[0071] The data acquisition hardware 101 is an intelligent hardware system integrating various sensing devices such as a high-definition camera, a pressure sensor, and a flow sensor. The pressure sensor is placed inside and around the simulated blood vessel to monitor the blood vessel pressure change in real time, and reflects the hemostasis effect with quantified data, providing a key basis for judging the impact of surgical operations on the restoration of blood vessel pressure. The flow sensor is embedded in the simulated blood circulation system to measure the liquid flow rate. By comparing the flow rate data before and after suture, the patency of the blood vessel and the quality of the suture can be effectively evaluated. The high-definition camera is placed in the surgical training simulation environment to collect various key data during the surgical operation in real time, including the movement trajectory of surgical instruments, the change of operation force, and the subtle deformation of the suture site.

[0072] The data processing device 102 is connected to various data acquisition hardware in the surgical training simulation device. It receives various information such as the surgical video images captured by the high-definition camera, the blood vessel pressure data collected by the pressure sensor, and the liquid flow rate data measured by the flow sensor. Through the pre-built professional discrimination software, these data are summarized, screened, cleaned, and deeply analyzed. The restoration and quantitative evaluation of the surgical training process are realized, and the performance levels of the surgical operator on multiple key indicators such as suture accuracy, hemostasis timeliness, and operation method standardization are judged, improving the training quality and effect of hepato-biliary artery suture hemostasis surgery.

[0073] The intelligent device terminal 103 is connected to data acquisition hardware and also establishes a communication link with a data processing device. It is used to obtain various types of information during the surgical training process in real time, such as the change in the force of the surgical instrument in contact with the tissue, the subtle movements when the suture needle passes through the blood vessel wall, and the visual image data of the entire operation process, etc., and quickly transmits these raw data to the data processing device for analysis. It can not only accurately point out the mistakes and deficiencies in the surgical operation, such as uneven suture spacing, improper ligation force, etc., but also quantitatively evaluate the proficiency and skill mastery of the trainer, and generate a detailed and targeted training report.

[0074] Combined with the above-mentioned noun explanations and implementation environments, the application scenarios of the embodiments of this application are described.

[0075] A method and training model for discriminating the results of hepatobiliary artery suture hemostasis surgery training provided by the embodiments of this application, by carefully arranging high-precision pressure sensors, displacement sensors, high-definition cameras and other devices at key positions around the surgical instrument, simulated artery tissue, etc., to collect various types of data during the surgical operation process in all directions and in real time. Subsequently, the rich data collected is transmitted to the intelligent data analysis platform, comprehensively considering multiple key indicators such as the uniformity of suture needle distance, the firmness of ligation, the timeliness and effectiveness of hemostasis, etc., and deeply matching and quantitatively evaluating with authoritative medical standards and excellent surgical case data, and finally discriminating the quality of the surgical training results. Exemplarily, a method and training model for discriminating the results of hepatobiliary artery suture hemostasis surgery training provided by the embodiments of this application can be applied to at least one of the following scenarios including but not limited to the following scenarios.

[0076] First, the method and training model for discriminating the results of hepatobiliary artery suture hemostasis surgery training are applied to teaching and training. In the surgical teaching laboratory of a medical college, the high-definition camera in the data acquisition hardware 101 is placed above and on the side of the surgical simulation table to capture every subtle movement of the trainee's operation in all directions. At the same time, the pressure sensor and flow sensor in the simulated blood vessel monitor the pressure change and liquid flow data of the blood vessel in real time, and these data are sent to the data processing device 102 through the wireless transmission module. The data processing device 102 performs image recognition and action decomposition on the video data captured by the camera, accurately analyzes key indicators such as the spacing of suture stitches, the suture angle, and the duration of the hemostasis operation, and combines the pressure and flow data to judge the hemostasis effect and patency of the blood vessel. The processed results are pushed to the intelligent device terminal 103 in real time, and the training situation can be viewed at any time, and the deficiencies in the operation can be pointed out, such as an overly wide suture needle distance may lead to poor blood vessel healing, or the pressure does not return to the normal range during hemostasis, etc.

[0077] Second, the application of the discrimination method for the training results of hepato-biliary artery suture hemostasis surgery and the training model in remote surgical skill assessment. In a surgical skill assessment, candidates are distributed in the simulated operating rooms of hospitals in different regions and are connected to the server of the assessment center through the network. When the candidates perform the hepato-biliary artery suture hemostasis surgery operation, the data acquisition hardware 101 in the examination room comprehensively collects the data of the surgical process and transmits it to the data processing device 102 of the assessment center in real time. The data processing device 102 quickly analyzes and scores the operation data of the candidates according to the preset strict assessment criteria, including multiple dimensions such as suture quality, hemostasis effect, and standardization of the operation process. The assessment experts remotely log in to the assessment system through their respective intelligent device terminals 103, and synchronously watch the live video of the candidates' operations and the real-time generated data reports. During the assessment process, the experts can record the highlights and problem points of the candidates' performances at any time on the intelligent device terminals 103. After the assessment, the system quickly generates a comprehensive assessment report according to the analysis results of the data processing device 102 and the evaluation opinions of the experts, and sends it to the candidates and the management departments of their affiliated hospitals.

[0078] In one embodiment, as Figure 1 shown, a discrimination method for the training results of hepato-biliary artery suture hemostasis surgery is provided, which may include the following steps:

[0079] Step S201, obtaining surgical training data related to the hepato-biliary artery suture hemostasis surgery; the surgical training data includes the physiological information of the tissue, the operation data of the surgical instruments, and the image information of the surgical operation area.

[0080] Collect the physiological information of the tissue, and the physiological information includes the degree of tissue deformation, internal pressure change, bleeding flow rate, and bleeding speed; these information can reflect the real-time physiological response of the tissue during the surgical operation and are crucial for evaluating the impact of the surgical operation on the tissue. Collect the operation data of the surgical instruments, including the displacement, angle, speed, applied force, and operation duration information of the instruments; these data can accurately present the proficiency and accuracy of the surgeon's manipulation of the instruments. Collect the image information of the surgical operation area to ensure that the morphology, color change, and detailed features of the suture site of the tissue can be clearly captured; these images provide an intuitive visual record of the surgical operation process and help analyze and evaluate the surgical effect.

[0081] Step S202, processing the collected surgical training data to obtain feature information related to the surgical training quality.

[0082] In the processing flow of training data for hepato-biliary artery suture hemostasis surgery, first, the physiological information of tissues, the operation data of surgical instruments, and the image information of the surgical operation area are processed to obtain physiological parameters, operation parameters, and image parameters. Subsequently, a data set is constructed based on each data parameter, and then the trained deep learning model is used to extract deep features from the data set to mine out feature information. Finally, the least absolute shrinkage and selection operator algorithm is used to assist in verifying the extracted feature information, further optimizing and improving the feature information to ensure that it can accurately reflect the key features and effects of surgical training.

[0083] Step S203: Input the feature information of the surgical training result to be judged into the trained discrimination model to obtain the discrimination result of the surgical training result; the discrimination result includes a grade evaluation for judging whether the surgical operation is qualified; among them, the discrimination model is trained by combining multiple machine learning classifiers.

[0084] The feature information of the surgical training result to be judged after a series of complex processing and extraction is input into the discriminant model that has been carefully trained in advance. This discriminant model is constructed and trained based on a large amount of historical surgical training data and corresponding professional evaluation results. After receiving the input feature information, the model performs in-depth operations and comparative analyses. The final discrimination result of the surgical training result covers multiple key aspects. For example, in terms of the rationality of tissue processing, it can clearly indicate whether the processing of the hepato-biliary artery tissue and surrounding tissues during the surgical operation follows the best medical practice principles, and whether there is excessive damage or improper handling; for the hemostasis effect, it will accurately judge whether the expected hemostasis standard has been achieved, whether it is complete hemostasis, partial hemostasis, or there is still a bleeding risk; in terms of the stability of surgical instrument operation, it can evaluate whether the operator's control of the instrument is stable and accurate, and whether there are unnecessary jitters or misoperations; regarding the accuracy of suture needle puncture and suturing, it will detail whether the needle insertion position, depth, angle of the suture needle, and the degree of suture line pulling are in line with the specification requirements. Through such a comprehensive and detailed discrimination result, targeted feedback and guidance can be provided to surgical trainees, so that they can improve and enhance purposefully, and then continuously improve the operation skill level of hepato-biliary artery suture hemostasis surgery.

[0085] The above-mentioned method for discriminating the training results of hepatobiliary artery suture hemostasis surgery, the training model, the computer device, and the storage medium first obtain relevant surgical training data including the physiological information of the tissue, the operation data of the surgical instrument, and the image information of the surgical operation area. Then, data processing technology is used to extract the collected training data, so as to refine the feature information closely related to the surgical training quality. Finally, the feature information of the surgical training result to be discriminated is input into the discrimination model trained by a variety of machine learning classifiers, and the discrimination result of the surgical training result including the grade evaluation of whether the surgical operation is qualified is output by this model, so as to realize the scientific and accurate evaluation and judgment of the surgical training effect.

[0086] In one embodiment, as Figure 2 shown, processing the collected surgical training data to obtain feature information related to the surgical training quality may include the following steps:

[0087] Step S301, perform noise reduction and filtering processing on the physiological information of the tissue and the operation data of the surgical instrument to obtain the processed physiological parameters and operation parameters.

[0088] Step S302, perform image enhancement processing, denoising, and grayscale processing on the image information of the surgical operation area to obtain the processed image parameters.

[0089] Step S303, construct a data set based on each data parameter in the physiological parameters, operation parameters, and image parameters to obtain the sample data to be processed.

[0090] Step S304, perform normalization and outlier removal operations on the sample data to be processed to obtain a feature data set.

[0091] Step S305, balance the data based on the feature data set using the synthetic minority over-sampling technique to obtain a balanced data set.

[0092] Step S306, use the trained deep learning model to extract features from the balanced data set to obtain the corresponding feature information; the feature information includes the rationality feature of tissue processing, the hemostasis effect feature, the stability feature for evaluating the operation stability of the surgical instrument, and the accuracy feature of suture needle puncture and suture.

[0093] The rationality characteristics include at least one of the degree of tissue deformation, the pressure distribution on the tissue, and the blood flow blockage of the tissue during the suturing process; the hemostasis effect characteristics include at least one of the area change rate of the bleeding area, the change in bleeding speed, and whether the bleeding finally stops; the stability characteristics include at least one of the force fluctuation range during the operation, the instrument displacement and the angle change rate; the accuracy characteristics include at least one of the deviation between the actual puncture position of the suture needle and the preset ideal puncture position, the bending angle change range of the suture needle, and the tension distribution of the suture line.

[0094] Step S307: Use the least absolute shrinkage and selection operator algorithm to perform auxiliary verification on the feature information to obtain updated feature information.

[0095] In the process of processing the training data of hepatobiliary artery suture hemostasis surgery, firstly, noise reduction and filtering are performed on the physiological information of the tissue and the operation data of the surgical instruments, and the interference components are effectively eliminated through advanced signal processing algorithms, so as to obtain accurate processed physiological parameters and operation parameters; then, image enhancement, denoising and grayscale processing are carried out on the image information of the surgical operation area, and professional image processing technology is used to highlight the key details of the image, eliminate noise and convert it into a grayscale image suitable for analysis, so as to obtain processed image parameters; then, a data set is constructed with the obtained data parameters, and multi-source data is integrated to form sample data to be processed; then, the sample data to be processed is normalized to unify the data to a specific value The range is selected and outliers are removed to generate a feature data set; based on the feature data set, the synthetic minority oversampling technology is used to deal with the data imbalance problem, increase the number of minority class samples, make the distribution of various types of data more balanced, and obtain a balanced data set; then, with the help of the trained deep learning model, the potential features of the balanced data set are deeply mined, and rich feature information including tissue treatment rationality features, hemostasis effect features, stability features for evaluating the stability of surgical instrument operation, and suture needle puncture and suture accuracy features are extracted; finally, the minimum absolute shrinkage and selection operator algorithm is used to assist in verifying the extracted feature information, and the key features are screened out and the feature weights are optimized according to the characteristics of the algorithm to obtain more accurate and effective updated feature information. These steps effectively improve the data quality and availability through fine processing of the data.

[0096] In one embodiment, if Figure 3 As shown in the figure, the trained deep learning model is used to extract features from the balanced data set to obtain the corresponding feature information, including:

[0097] Step S401, the data in the balanced data set is processed using a recursive feature elimination method to obtain different feature subsets.

[0098] Step S402: Integrate different feature subsets using the multi-layer perceptron algorithm to obtain a comprehensive feature set that combines all data parameters.

[0099] Step S403: Perform k-fold cross-validation on the integrated comprehensive feature set to obtain corresponding performance metrics; the performance metrics include the rationality metric for tissue processing, the hemostasis effect metric, the stability metric for evaluating the operating stability of surgical instruments, and the accuracy metric for suture needle puncture and suturing.

[0100] Step S404: Calculate the performance metrics to evaluate the contribution degree of each performance metric to the discriminant model, and obtain the feature evaluation value.

[0101] Step S405: Screen the feature evaluation values to obtain the feature information that meets the conditions of the preset evaluation metrics.

[0102] In this embodiment, first, the recursive feature elimination method is used to process the balanced data set to obtain feature subsets, then the multi-layer perceptron algorithm is used to integrate them into a comprehensive feature set, and then k-fold cross-validation is performed on it to obtain performance metrics in multiple aspects such as tissue processing and hemostasis effect. Subsequently, the contribution degree of each performance metric to the discriminant model is calculated and evaluated to obtain the feature evaluation value. Finally, the feature information that meets the conditions of the preset evaluation metrics is screened out. It can accurately extract key data, effectively improve the quality and representativeness of features, make the obtained feature information highly consistent with the requirements of the discriminant model, and thus improve the discriminant accuracy of the model for the training results of hepatobiliary artery suture hemostasis surgery, helping to improve the level of surgical training.

[0103] In one of the embodiments, calculating the performance metrics to evaluate the contribution degree of each performance metric to the discriminant model and obtaining the feature evaluation value includes:

[0104] Step S501: Calculate based on the rationality metric for tissue processing using a formula to obtain the corresponding rationality feature evaluation value.

[0105]

[0106] Where, S t represents the rationality feature evaluation value, k represents the number of monitoring points, l represents the actual tissue deformation degree of the th monitoring point, T n represents the normal allowable tissue deformation degree, F l represents the actual tissue force of the th monitoring point, F n represents the safe force range that the tissue can withstand, V l represents the tissue blood flow velocity change rate of the th monitoring point, V n represents the acceptable range of blood flow velocity change rate.

[0107] Step S502: Calculate using a formula based on the hemostasis effect index of tissue processing to obtain the corresponding hemostasis effect characteristic evaluation value.

[0108]

[0109] Among them, S h represents the hemostasis effect characteristic evaluation value, t represents the bleeding stop time, and t s represents the preset standard hemostasis time, A b represents the final bleeding area, and A a represents the maximum acceptable bleeding area.

[0110] Step S503: Calculate using a formula based on the stability index for evaluating the operating stability of the surgical instrument to obtain the corresponding stability characteristic evaluation value.

[0111]

[0112] Among them, S a represents the surgical instrument operating stability score, n represents the number of sampling points, and F i , D i , A i respectively represent the force, displacement, and angle values at the i-th sampling point, represents the average value of the force, displacement, and angle, and F max , D max , A max represent the maximum values of the force, displacement, and angle.

[0113] Step S504: Calculate using a formula based on the accuracy index of suture needle puncture and suture to obtain the corresponding accuracy characteristic evaluation value.

[0114]

[0115] Among them, S c represents the suture needle puncture accuracy characteristic evaluation value, m represents the number of punctures, and P j represents the deviation distance between the actual position and the target position of the i-th puncture, and P t represents the preset maximum acceptable deviation distance, B j represents the actual bending angle of the suture needle during the i-th puncture, and B t represents the preset ideal bending angle range.

[0116] In the quantitative analysis of the relevant indicators of the training results of hepato-biliary artery suture hemostasis surgery, according to the rationality index of tissue treatment, substituting into a specific formula, comprehensively considering factors such as the tissue deformation degree, force, and blood flow velocity change rate at the monitoring points, the rationality characteristic evaluation value is calculated; based on the hemostasis effect index, combined with parameters such as the bleeding stop time and the final bleeding area, the hemostasis effect characteristic evaluation value is calculated through the corresponding formula; using the stability index of the surgical instrument operation stability and the given formula, the stability characteristic evaluation value is calculated according to the force, displacement, and angle values at the sampling points; according to the accuracy index and the corresponding formula of the suture needle puncture and suture, the accuracy characteristic evaluation value is calculated based on the puncture deviation distance and the bending angle, etc. These steps provide a quantitative basis for comprehensively evaluating the surgical training effect.

[0117] In one embodiment, as Figure 4 shown, the characteristic information of the surgical training result to be discriminated is input into the trained discrimination model to obtain the discrimination information of the surgical training result, including:

[0118] Step S601, data label each characteristic information according to the fuzzy comprehensive evaluation method to obtain the characteristic category parameters corresponding to each characteristic information.

[0119] Step S602, perform missing value processing on the processed characteristic category parameters to obtain the missing value analysis result.

[0120] Among them, when the missing value ratio of all characteristic category parameters is less than 10%, the missing value analysis result is abnormal. Then, for the missing characteristic category parameters, the mode filling method is used for data filling, and for the continuous characteristic category parameters, the random forest algorithm is used for interpolation.

[0121] When the missing value ratio of all characteristic category parameters is greater than or equal to 10%, the missing value analysis result is normal, and the characteristic category parameters are kept unchanged.

[0122] Step S603, update the characteristic category parameters based on the missing value analysis result to obtain the updated characteristic category parameters.

[0123] Step S604, input each updated characteristic category parameter into the trained discrimination model to obtain the discrimination information of the surgical training result.

[0124] Step S605, perform grade classification on the discrimination information based on the grade evaluation formula to obtain the discrimination result for classifying and discriminating the surgical training result.

[0125] The discrimination result includes a grade evaluation on whether the surgical operation is qualified. The grade of whether the surgical operation is qualified is divided into four levels: excellent, good, qualified, and unqualified. When the discrimination result is unqualified, the specific reasons for the unqualified are further output, and the reasons include at least one of suture operation errors, unstable instrument operation, improper tissue handling, or incomplete hemostasis.

[0126] First, label each characteristic information according to the fuzzy comprehensive evaluation method to obtain characteristic category parameters. Then, perform missing value processing on these parameters according to the missing value ratio to obtain an analysis result. Decide whether to fill or retain the operation based on this to update the parameters. Then, input the updated parameters into the trained discrimination model to obtain discrimination information. Finally, use the grade evaluation formula to classify the discrimination information to obtain the discrimination result of the surgical training result, thereby completing the comprehensive evaluation of the training result of the hepatobiliary artery suture hemostasis surgery.

[0127] In one of the embodiments, as Figure 5 shown, the trained discrimination model is obtained through the following steps, including:

[0128] Step S701, use a machine learning classifier to train the model to obtain a discrimination classifier. Among them, the machine learning classifier is at least one of Gaussian Naive Bayes, neural network, ridge regression, and linear logistic regression.

[0129] Step S702, use the grid search method to train each discrimination classifier to obtain a pre-trained discrimination model.

[0130] Step S703, perform normalization processing on the characteristic information based on the model trained with historical data to obtain a characteristic classification sample set.

[0131] Step S704, process each characteristic classification sample set until the characteristic evaluation values corresponding to each characteristic information meet the preset evaluation index conditions to obtain an equalized characteristic classification sample set.

[0132] Step S705, use each characteristic information in the equalized characteristic classification sample set as the input of the current pre-trained discrimination model with the initial value determined, and use the discrimination information corresponding to each characteristic information as the expected output of the current pre-trained discrimination model with the initial value determined.

[0133] Step S706, construct a loss function for the actual output and the expected output of the current discrimination model, perform backpropagation based on the loss function, and modify the undetermined network parameters of the current pre-trained discrimination model until the loss function meets the set requirements to complete the training of the current pre-trained discrimination model.

[0134] First, at least one machine learning classifier such as Gaussian Naive Bayes and neural network is used for training to obtain a discriminant classifier. Then, the grid search method is used for further training to obtain a pre-trained discriminant model. After that, the feature information is normalized into a feature classification sample set based on the model trained with historical data, and processed until it meets the conditions of the preset evaluation index to obtain a balanced sample set. Then, its feature information is used as the input of the pre-trained discriminant model, and the discriminant information is used as the expected output. Finally, a loss function is constructed and backpropagated to modify the undetermined network parameters until the set requirements are met, and the training of the discriminant model is completed. It can effectively improve the accuracy and reliability of the discriminant model, enable it to better process the training data of hepatobiliary artery suture hemostasis surgery, accurately discriminate the results of surgical training, and help medical staff improve their surgical skills.

[0135] In one embodiment, the discriminant information is classified by level based on the level evaluation formula to obtain the discriminant result for classifying and discriminating the surgical training result, including:

[0136] The following formula is used to calculate the discriminant information to evaluate the influence of multiple key factors on the surgical effect and obtain the comprehensive score of the surgical training result.

[0137]

[0138] Among them, S represents the comprehensive score of the surgical training result, n represents the number of feature category parameters affecting the surgical training result, w i represents the weight of the i-th feature category parameter, S i represents the feature evaluation value of the i-th feature category parameter, f i (S i ) represents the score conversion function for the i-th feature category parameter, which converts the feature evaluation value into a score within the range of [0, 1].

[0139] Based on the preset scoring interval, the comprehensive score of the surgical training result is judged to obtain the discriminant result of classification and discrimination.

[0140] The discriminant information is calculated through a specific formula. According to the number, weight, feature evaluation value, and score conversion function of each feature category parameter affecting the surgical training result, the comprehensive score of the surgical training result is obtained. Then, this comprehensive score is judged according to the preset scoring interval, so as to obtain the discriminant result of classification and discrimination, thereby realizing the quantitative evaluation and classification determination of the surgical training effect. By using this formula and subsequent judgment steps, the influence of multiple key factors on the surgical effect can be systematically integrated, the complex surgical training situation can be quantified into an intuitive comprehensive score, and the discriminant result can be accurately obtained according to the preset scoring interval, providing an objective, scientific, and efficient method for evaluating the surgical training effect and strongly guiding subsequent training improvement and skill improvement.

[0141] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the indications of the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps has no strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or steps or stages in other steps.

[0142] In one embodiment, as Figure 6 shown, a training model for discriminating the training results of hepato-biliary artery suture hemostasis surgery is provided. The training model includes:

[0143] A data acquisition module 801, configured to acquire surgical training data related to hepato-biliary artery suture hemostasis surgery; the surgical training data includes physiological information of tissues, operation data of surgical instruments, and image information of the surgical operation area.

[0144] A data processing module 802, configured to process the acquired surgical training data to obtain feature information related to the quality of surgical training.

[0145] A model training module 803, configured to input the feature information of the surgical training results to be discriminated into the trained discrimination model to obtain the discrimination result of the surgical training results; the discrimination result includes a grade evaluation for judging whether the surgical operation is qualified; wherein, the discrimination model is trained by combining multiple machine learning classifiers.

[0146] In the hepato-biliary artery suture hemostasis surgery training result discrimination system, the data acquisition module 801 plays a key role. It can accurately acquire surgical training data closely related to surgical training, including physiological information of tissues, operation data of surgical instruments, and image information of the surgical operation area. Then, the data processing module 802 performs a series of complex and orderly processing on the acquired data. Through professional algorithms and technical means, it finally extracts feature information closely related to the quality of surgical training. Subsequently, the model training module 803 inputs the feature information of the surgical training results to be discriminated into the discrimination model that has been carefully trained and combines multiple machine learning classifiers. After in-depth operation and analysis of the model, the discrimination result of the surgical training results, including the grade evaluation for judging whether the surgical operation is qualified, is obtained, providing an important basis for the evaluation and improvement of surgical training.

[0147] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of a method for discriminating the training result of a hepatobiliary artery suture hemostasis operation and a training model as described above are implemented.

[0148] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0149] For the device embodiment, since it basically corresponds to the method embodiment, the relevant parts can be referred to the partial description of the method embodiment. The device embodiments described above are only illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. A person of ordinary skill in the art can understand and implement it without creative work.

[0150] The above embodiments only represent several implementation manners of the embodiments of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the embodiments of the application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several deformations and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A discriminant method for the training results of hepatobiliary artery suture hemostasis surgery, characterized in that, The method includes: Obtaining surgical training data related to hepato-biliary artery suture hemostasis surgery training; the surgical training data includes physiological information of tissues, operation data of surgical instruments, and image information of the surgical operation area; Processing the collected surgical training data to obtain feature information related to the quality of surgical training; Inputting the feature information of the surgical training result to be discriminated into a trained discrimination model to obtain a discrimination result of the surgical training result; the discrimination result includes a grade evaluation for judging whether the surgical operation is qualified; wherein, the discrimination model is trained by combining multiple machine learning classifiers.

2. The method according to claim 1, wherein The processing the collected surgical training data to obtain feature information related to the quality of surgical training includes: Performing noise reduction and filtering processing on the physiological information of tissues and the operation data of surgical instruments to obtain processed physiological parameters and operation parameters; Performing image enhancement processing, denoising, and grayscale processing on the image information of the surgical operation area to obtain processed image parameters; Constructing a data set based on each data parameter in the physiological parameters, operation parameters, and image parameters to obtain sample data to be processed; Performing normalization and outlier removal operations on the sample data to be processed to obtain a feature data set; Balancing the data based on the feature data set using the Synthetic Minority Over-sampling Technique (SMOTE) to obtain a balanced data set; Using a trained deep learning model to extract features from the balanced data set to obtain corresponding feature information; the feature information includes the rationality feature of tissue processing, the hemostasis effect feature, the stability feature for evaluating the operation stability of surgical instruments, and the accuracy feature of suture needle puncture and suture; Using the Least Absolute Shrinkage and Selection Operator (LASSO) algorithm to perform auxiliary verification on the feature information to obtain updated feature information.

3. The method according to claim 2, wherein The using a trained deep learning model to extract features from the balanced data set to obtain corresponding feature information includes: Processing the data in the balanced data set using the Recursive Feature Elimination (RFE) method to obtain different feature subsets; Using a Multi-Layer Perceptron (MLP) algorithm to integrate different feature subsets to obtain a comprehensive feature set that combines all the data parameters; Performing k-fold cross-validation on the integrated comprehensive feature set to obtain corresponding performance indicators; the performance indicators include the rationality indicator of tissue processing, the hemostasis effect indicator, the stability indicator for evaluating the operation stability of surgical instruments, and the accuracy indicator of suture needle puncture and suture; Calculating the performance indicators to evaluate the contribution degree of each performance indicator to the discrimination model to obtain feature evaluation values; Screening the feature evaluation values to obtain feature information that meets the conditions of the preset evaluation indicators.

4. The method according to claim 3, wherein The calculating the performance indicators to evaluate the contribution degree of each performance indicator to the discrimination model to obtain feature evaluation values includes: Calculating based on the rationality indicator of tissue processing using a formula to obtain a corresponding rationality feature evaluation value; Among them, S t represents the rationality characteristic evaluation value, k represents the number of monitoring points, l represents the actual deformation degree of the tissue at the l-th monitoring point, T n represents the normal allowable tissue deformation degree, F l represents the actual force on the tissue at the l-th monitoring point, F n represents the safe force range that the tissue can withstand, V l represents the tissue blood flow velocity change rate at the l-th monitoring point, V n represents the acceptable range of blood flow velocity change rate; Calculating based on the hemostasis effect indicator of tissue processing using a formula to obtain a corresponding hemostasis effect feature evaluation value; Among them, S h represents the evaluation value of the hemostatic effect characteristics, t represents the bleeding stop time, t s represents the preset standard hemostatic time, A b represents the final bleeding area, A a represents the maximum acceptable bleeding area; The stability index for evaluating the operating stability of the surgical instrument is calculated using a formula to obtain a corresponding stability characteristic evaluation value; Among them, S a represents the operation stability score of the surgical instrument, n represents the number of sampling points, and F i , D i , A i respectively represent the force, displacement, and angle values at the i-th sampling point. represents the average value of the force, displacement, and angle, and F max , D max , A max represent the maximum values of the force, displacement, and angle. The accuracy index for suture needle puncture and suture is calculated using a formula to obtain a corresponding accuracy characteristic evaluation value; Among them, S c represents the evaluation value of the suture needle puncture accuracy feature, m represents the number of punctures, and P j represents the deviation distance between the actual position and the target position of the nth puncture, and P t represents the preset maximum acceptable deviation distance, B j represents the actual bending angle of the suture needle during the nth puncture, and B t represents the preset ideal bending angle range.

5. The method according to claim 1, wherein Inputting the characteristic information of the surgical training result to be discriminated into the trained discrimination model to obtain discrimination information on the surgical training result, including: Data-labeling each of the characteristic information according to the fuzzy comprehensive evaluation method to obtain a characteristic category parameter corresponding to each of the characteristic information; Performing missing value processing on the processed characteristic category parameters to obtain a missing value analysis result; Wherein, when the missing value ratio of all the characteristic category parameters is less than 10%, the missing value analysis result is abnormal, and for the missing characteristic category parameters, the method of filling with the mode is used for data filling, and for continuous characteristic category parameters, the random forest algorithm is used for interpolation; When the missing value ratio of all the characteristic category parameters is greater than or equal to 10%, the missing value analysis result is normal, and the characteristic category parameters are kept unchanged; Updating the characteristic category parameters based on the missing value analysis result to obtain updated characteristic category parameters; Inputting each of the updated characteristic category parameters into the trained discrimination model to obtain discrimination information on the surgical training result; Performing grade classification on the discrimination information based on a grade evaluation formula to obtain a discrimination result for classifying and discriminating the surgical training result.

6. The method according to claim 5, wherein The trained discrimination model is obtained through the following steps, including: Using a machine learning classifier for model training to obtain a discrimination classifier; wherein, the machine learning classifier is at least one of Gaussian Naive Bayes, neural network, ridge regression, and linear logistic regression; Using a grid search method to train each of the discrimination classifiers to obtain a pre-trained discrimination model; Performing normalization processing on the characteristic information based on a model trained with historical data to obtain a characteristic classification sample set; Processing each of the characteristic classification sample sets until the characteristic evaluation values corresponding to each of the characteristic information meet the preset evaluation index conditions to obtain an equalized characteristic classification sample set; Taking each of the characteristic information in the equalized characteristic classification sample set as the input of the current pre-trained discrimination model with the initial value determined, and taking the discrimination information corresponding to each of the characteristic information as the expected output of the current pre-trained discrimination model with the initial value determined; Constructing a loss function for the actual output and the expected output of the current discrimination model, performing backpropagation based on the loss function, and modifying the undetermined network parameters of the current pre-trained discrimination model until the loss function meets the set requirements to complete the training of the current pre-trained discrimination model.

7. The method according to claim 5, characterized in that, The performing grade classification on the discrimination information based on a grade evaluation formula to obtain a discrimination result for classifying and discriminating the surgical training result, including: Calculating the discrimination information using the following formula to evaluate the influence of multiple key factors on the surgical effect to obtain a comprehensive score of the surgical training result; Among them, S represents the comprehensive score of the surgical training result, n represents the number of feature category parameters affecting the surgical training result, and w i represents the weight of the i-th feature category parameter, and S i represents the feature evaluation value of the i-th feature category parameter, and f i (S i ) represents the score conversion function for the i-th feature category parameter, which converts the feature evaluation value into a score within the range of [0, 1]; Based on a preset scoring range, judge the comprehensive score of the surgical training result to obtain the discrimination result of classification discrimination.

8. A training model for discriminating the training results of a hepatobiliary artery suture hemostasis operation, characterized in that, The training model includes: A data acquisition module, configured to acquire surgical training data related to hepatobiliary artery suture hemostasis surgical training; the surgical training data includes physiological information of tissues, operation data of surgical instruments, and image information of the surgical operation area; A data processing module, configured to process the acquired surgical training data to obtain feature information related to the quality of surgical training; A model training module, configured to input the feature information of the surgical training result to be discriminated into a trained discrimination model to obtain the discrimination result of the surgical training result; the discrimination result includes a grade evaluation for judging whether the surgical operation is qualified; wherein, the discrimination model is trained by combining multiple machine learning classifiers.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 7.

Images