Neurosurgical operation risk intelligent assessment system

Abstract

The invention discloses an intelligent assessment system for neurosurgery operation risks. The system comprises a preoperative data collection module, a data optimization module, an operation planning and design module, a preoperative nursing scheme optimization module and a preoperative risk comprehensive prediction module. The invention belongs to the technical field of operation risk assessment, and particularly relates to a neurosurgery operation risk intelligent assessment system, which performs neurosurgery operation risk intelligent assessment by combining operation planning design and preoperative nursing scheme optimization, can effectively combine specific requirements of an operation for nursing, and can effectively improve the operation risk of the operation. The implementation of the operation can be analyzed according to the specific condition of the patient, so that the relevance between preoperative analysis and operation implementation is improved, and analysis and evaluation of numerous and jumbled patient data are also realized; a difference optimization reinforcement learning method integrating improved convolution feature extraction is adopted to carry out operation planning design; an improved quantum convolution long and short term neural network combining risk framework analysis and random forest is adopted to optimize a preoperative nursing scheme.

Description

Technical Field

[0001] The present invention belongs to the technical field of surgical risk assessment, and in particular relates to an intelligent assessment system for neurosurgery risk. Background Art

[0002] The neurosurgery risk intelligent assessment system uses artificial intelligence technology to intelligently predict and assess the patient's risk during neurosurgery by integrating the patient's preoperative assessment data, surgical planning information, and historical case data. The system calculates a personalized surgical risk score by analyzing multi-dimensional health characteristics, such as the patient's age, underlying diseases, physical signs, imaging data, and surgery type, combined with deep learning or machine learning algorithms. Its main function is to provide clinicians with accurate risk warnings, assist in decision-making, and optimize preoperative care plans, thereby improving surgical safety, reducing the incidence of complications, and promoting patient recovery after surgery.

[0003] However, in the existing intelligent assessment system of neurosurgical risk, there is a problem that the existing intelligent assessment method or system of neurosurgical risk mainly relies on the combined perspective of preoperative and postoperative assessment and prediction. Since neurosurgical surgery involves more complex nursing conditions before surgery and patients also have a variety of preoperative nursing risks, the existing methods for comprehensive preoperative surgical risk assessment of neurosurgical surgery are not detailed enough and the referenceability is not high enough, and the connection between preoperative analysis and the implementation of the surgery itself is not enough. Technical problems; in the existing neurosurgical surgery planning and design methods, there is a problem that the preoperative data involved in neurosurgical surgery has multi-source and multi-modal characteristics, and the complex heterogeneous data increases the difficulty of constructing the machine model of the surgical steps, and the existing methods are difficult to effectively connect the multi-source heterogeneous preoperative data to the multi-objective surgical planning and design; in the existing optimization of neurosurgical preoperative care plans, there is a technical problem that the preoperative care combined with surgical planning and design involves more complex time trend changes in the health status of patients and the characteristics of micro-lesions in medical data, and the existing methods are difficult to process and fine-grainedly analyze these complex characteristics at the same time. Summary of the invention

[0004] The technical solution adopted by the present invention is as follows: A neurosurgery risk intelligent assessment system provided by the present invention includes a preoperative data collection module, a data optimization module, a surgery planning and design module, a preoperative nursing program optimization module and a preoperative risk comprehensive prediction module;

[0005] The preoperative data collection module is used for preoperative data collection, obtains original data of neurosurgery risk assessment through preoperative data collection, and sends the original data of neurosurgery risk assessment to the data optimization module;

[0006] The data optimization module is used for data optimization, and obtains optimized surgical risk assessment data through data optimization, and sends the optimized surgical risk assessment data to the surgical planning and design module, the preoperative nursing program optimization module and the preoperative risk comprehensive prediction module;

[0007] The surgical planning and design module is used for surgical planning and design, obtains surgical planning and design reference data through surgical planning and design, and sends the surgical planning and design reference data to the preoperative nursing program optimization module;

[0008] The preoperative care program optimization module is used to optimize the preoperative care program, obtain preoperative care program reference data through the preoperative care program optimization, and send the preoperative care program reference data to the preoperative risk comprehensive prediction module;

[0009] The preoperative risk comprehensive prediction module is used for preoperative risk comprehensive prediction. Through preoperative risk comprehensive prediction, reference data for neurosurgery risk assessment is obtained, and intelligent neurosurgery risk assessment is performed.

[0010] Furthermore, the preoperative data collection is used to collect multimodal data of neurosurgery patients, specifically, to obtain original data of neurosurgery risk assessment from the medical system through data collection;

[0011] The original data of neurosurgery risk assessment specifically includes basic patient information data, clinical examination data, physiological characteristics data, neurological function examination data, patient symptom data and preoperative assessment data;

[0012] The patient's basic information data specifically includes basic personal information and medical history information; the basic personal information includes name, gender, age, height and weight; the medical history information includes disease history, allergy history, family medical history and surgical history information;

[0013] The clinical examination data specifically refers to vital signs data and laboratory examination data; the vital signs data specifically includes heart rate, respiratory rate, blood pressure, body temperature and blood oxygen saturation data; the laboratory examination data specifically includes blood routine, blood sugar, liver and kidney function, coagulation function, blood lipids and inflammatory index data;

[0014] The physiological characteristic data specifically include electrocardiogram data, vital capacity data, urine routine data and functional imaging characteristic data; the functional imaging characteristic data specifically include magnetic resonance imaging characteristics and cerebral angiography characteristic data;

[0015] The neurological function examination data specifically includes neurological sign data and neurological imaging assessment data; the neurological sign data specifically includes motor examination data, sensory examination data, reflex examination data and cognitive ability examination data; the neurological imaging assessment data specifically refers to electroencephalogram assessment data;

[0016] The patient symptom data specifically includes patient-reported symptom data and symptom quantitative assessment data;

[0017] The preoperative evaluation data specifically include comorbidity screening data, anesthesia risk grading assessment data and body mass index assessment data.

[0018] Furthermore, the data optimization is used to clean, fuse and optimize the original data, specifically to optimize the original data of the neurosurgery risk assessment to obtain optimized surgical risk assessment data, and specifically includes the following steps: data cleaning optimization, data standardization optimization, data fusion optimization, feature optimization and data integration optimization;

[0019] The data cleaning and optimization specifically includes performing data deduplication, missing value processing and outlier processing operations on the original data of the neurosurgery risk assessment to obtain cleaned and optimized data; the missing value processing specifically includes performing missing value processing by using mean filling; the outlier processing specifically includes performing anomaly detection on the physiological characteristic data and performing outlier processing by using the box plot method;

[0020] The data standardization optimization specifically includes performing a data standardization operation on the data in the cleaning optimization data, specifically using Z-score standardization to obtain standardized data;

[0021] The data fusion optimization specifically involves extracting features from artificial image data and performing physiological data correlation analysis to obtain comprehensive patient health data, and performing data fusion optimization by combining multi-source and multi-modal data, and obtaining fused and optimized data by aligning time series data;

[0022] The feature optimization is specifically to perform statistical feature selection and principal component analysis feature dimension reduction operations on the feature data in the fused optimization data to obtain feature optimized data;

[0023] The data integration optimization specifically includes performing data consistency check and time series alignment on the feature optimization data and the fusion optimization data to obtain optimized surgical risk assessment data;

[0024] The optimized surgical risk assessment data specifically includes basic patient information characteristics, comprehensive patient health characteristics and preoperative assessment characteristics;

[0025] The patient's basic information characteristics and comprehensive patient health characteristics are used for surgical planning and design;

[0026] The comprehensive patient health characteristics and preoperative assessment characteristics are used to optimize the preoperative care plan.

[0027] Furthermore, the surgical planning design is used to perform personalized surgical path design, instrument recommendation and surgical difficulty assessment according to the patient's specific symptoms. Specifically, based on the optimized surgical risk assessment data, a differential optimization reinforcement learning method of integrated improved convolution feature extraction is used to perform surgical planning design to obtain surgical planning design reference data, which specifically includes the following steps: constructing a heterogeneous fusion integrated improved convolution layer, constructing a reinforcement learning surgical path planning layer, constructing differential evolution optimized surgical instrument planning, surgical planning design model training and surgical planning design;

[0028] The heterogeneous fusion integrated improved convolution layer is constructed to extract surgical planning features, specifically, a standard convolutional neural network and a pre-trained heterogeneous integrated convolutional image feature extraction model are constructed, multi-source heterogeneous data feature extraction is performed through classification feature extraction, and fusion surgical planning feature data is obtained through feature fusion;

[0029] The feature fusion specifically refers to fusing the optimization features of the surgical planning numerical data and the optimization features of the surgical planning impact data to obtain fused surgical planning feature data;

[0030] The reinforcement learning surgical path planning layer is constructed to optimize the design of the specific surgical implementation plan, specifically, based on the fused surgical planning feature data, a reinforcement learning model is constructed to optimize the surgical planning path decision, and the reinforcement learning surgical planning output is obtained through reinforcement learning environment construction and deep Q value learning;

[0031] The reinforcement learning environment construction is specifically to design a parameter environment in combination with the surgical planning design, which specifically includes the definition of the parameter environment, the definition of the state parameters, the definition of the action parameters and the definition of the reward function;

[0032] The parameter environment definition specifically refers to simulating the process environment of neurosurgery, including state parameters, action parameters and reward functions;

[0033] The state parameter definition specifically defines the surgical stages and elements during the surgical process, including imaging characteristics, patient physiological characteristics, and surgical step characteristics;

[0034] The action parameter definitions specifically include tumor removal actions, precision cutting actions, hemostasis operation actions, neurovascular repair suturing actions, tissue repair actions, and intraoperative monitoring actions;

[0035] The reward function definition specifically refers to defining it in combination with the surgical success rate, safety, surgical time and surgical risk, and constructing a reward function to perform iterative deep Q-value learning to obtain a reinforcement learning surgical planning output;

[0036] The calculation formula of the reward function is:

[0037] ;

[0038] In the formula, R t is the reward function, w1 is the weight of the surgical success rate, Success(·) is the surgical success rate function, A t is the action parameter, where t is the iteration index, w2 is the surgical safety weight, Safety(·) is the surgical safety function, S t is the state parameter, w3 is the surgery time weight, Time(·) is the surgery success rate function, w4 is the surgery risk weight, Risk(·) is the surgery risk function;

[0039] The iterative deep Q-value learning specifically adopts a Q-value update algorithm combined with policy gradient optimization to perform reinforcement learning optimization;

[0040] The reinforcement learning surgery planning output specifically includes optimal reference steps for surgery, optimal surgery operation selection, reference surgery risk, and surgery time estimation;

[0041] The construction of differential evolution optimization surgical instrument planning is used to optimize the selection of surgical instruments in surgical planning, specifically to construct a standard differential evolution algorithm, initialize a candidate surgical instrument combination, and perform mutation and crossover operations to optimize the surgical planning instrument selection, thereby obtaining surgical planning instrument selection reference data;

[0042] The surgical planning design model training is specifically to train the surgical planning design model through the heterogeneous fusion integrated improved convolution layer, the reinforcement learning surgical path planning layer and the differential evolution optimized surgical instrument planning, and to evaluate the surgical difficulty by combining the surgical path planning and the surgical instrument planning to obtain the surgical planning design model Model OG ;

[0043] The surgical planning design is specifically to use the surgical planning design model Model OG , performing surgical planning and design based on the optimized surgical risk assessment data to obtain surgical planning and design reference data;

[0044] The surgical planning and design reference data specifically includes surgical path reference data, surgical instrument reference data and surgical difficulty assessment data;

[0045] The surgical path reference data specifically includes optimal surgical reference steps, optimal surgical operation selection, surgical reference risks and surgical time estimation;

[0046] The surgical difficulty assessment data specifically includes difficulty quantitative scoring valuation data and difficulty grading data.

[0047] Furthermore, the preoperative care plan optimization is used to generate a preoperative care plan for patients suitable for surgery, specifically based on the optimized surgical risk assessment data and the surgical planning design reference data, using an improved quantum convolutional long-short-term neural network that combines risk framework analysis and random forests to optimize the preoperative care plan and obtain preoperative care plan reference data, specifically including the following steps: constructing a preoperative care risk factor framework model, constructing a quantum convolutional neural network, constructing a long-short-term memory neural network, integrating feature outputs, constructing a random forest decision subnet, training a preoperative care plan optimization model, and optimizing the preoperative care plan;

[0048] The constructing of the preoperative nursing risk factor framework model specifically includes introducing a nursing risk factor framework model composed of preoperative assessment characteristics and surgical planning design reference data, and performing optimization prediction of the preoperative nursing plan based on the nursing risk factor framework model to obtain preoperative nursing risk assessment data;

[0049] The constructing of the quantum convolutional neural network specifically constructs a standard quantum convolutional neural network including a quantum convolutional layer and a quantum pooling layer, and extracts preoperative care plan features and predicts drug management in combination with the optimized surgical risk assessment data to obtain preoperative care drug management features;

[0050] The feature output integration is specifically to perform weighted integration fusion on the output features of the preoperative care risk factor framework model, quantum convolutional neural network and long short-term memory neural network to obtain the optimized feature data of the comprehensive preoperative care plan;

[0051] The construction of the random forest decision subnet specifically includes taking the optimized feature data of the comprehensive preoperative nursing plan as the input of the random forest ensemble learning model, and constructing a standard random forest model for comprehensive optimization of the nursing plan to obtain an optimized output of the nursing plan;

[0052] The preoperative care program optimization model training is specifically to perform preoperative care program optimization model training through the preoperative care risk factor framework model, the quantum convolutional neural network, the long short-term memory neural network, the feature output integration and the random forest decision subnet to obtain the preoperative care program optimization model Model HQLR ;

[0053] The preoperative care program optimization is specifically to use the preoperative care program optimization model ModelHQLR , optimizing the preoperative care plan based on the optimized surgical risk assessment data and the surgical planning design reference data to obtain preoperative care plan reference data;

[0054] The preoperative care plan reference data specifically includes a preoperative drug management reference plan, a preoperative fasting arrangement reference plan, a preoperative health status trend monitoring reference plan and preoperative care risk assessment data.

[0055] Furthermore, the comprehensive preoperative risk prediction is used to perform a comprehensive surgical risk assessment, specifically, based on the optimized surgical risk assessment data, the surgical planning design reference data and the preoperative care program reference data, an integrated learning deep neural network method is used to perform a comprehensive preoperative risk prediction to obtain reference data for neurosurgery risk assessment;

[0056] The neurosurgery risk assessment reference data specifically includes a neurosurgery planning and design assessment reference plan and a neurosurgery preoperative care plan assessment reference plan.

[0057] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0058] (1) In the existing intelligent risk assessment system for neurosurgery, the existing intelligent risk assessment methods or systems for neurosurgery mainly rely on the combined perspectives of preoperative and postoperative assessment and prediction. However, since neurosurgery involves relatively complex nursing conditions before surgery and patients also have a variety of preoperative nursing risks, the existing methods for comprehensive preoperative risk assessment of neurosurgery are not detailed enough and the referenceability is not high enough, and the correlation between preoperative analysis and the implementation of the surgery itself is also insufficient. This solution creatively adopts the combination of surgical planning and design and preoperative nursing plan optimization to conduct intelligent risk assessment of neurosurgery. By predicting the risks and designing the surgical planning itself, and also conducting a joint analysis of the preoperative nursing of neurosurgery, it can effectively combine the specific needs of the surgery for nursing care, and can also analyze the implementation of the surgery according to the specific conditions of the patient, thereby improving the correlation between preoperative analysis and surgical implementation, and also realizing the analysis and evaluation of complex patient data;

[0059] (2) In the existing neurosurgical surgery planning and design methods, there are technical problems that the preoperative data involved in neurosurgical surgery has the characteristics of multi-source and multi-modality. At the same time, the complex heterogeneous data increases the difficulty of constructing the machine model of the surgical steps. The existing methods are difficult to effectively connect the multi-source heterogeneous preoperative data to the multi-objective surgical planning and design. This solution creatively adopts the differential optimization reinforcement learning method of integrated improved convolutional feature extraction to carry out surgical planning and design. Through the integration and improvement of multiple convolutional networks, the processing capability of multi-source and multi-structure data is optimized. At the same time, the differential algorithm and reinforcement learning are combined to design and predict surgical instruments and surgical steps, which simultaneously improves the preoperative foresight and knowability of neurosurgical surgery, and provides a more understandable and usable intelligent solution for the concrete decomposition of surgical risks;

[0060] (3) In the optimization of existing neurosurgery preoperative care plans, there is a technical problem that preoperative care combined with surgical planning and design involves relatively complex time trend changes in patients' health conditions and micro-lesion characteristics in medical data. Existing methods are difficult to process and analyze these complex characteristics at the same time. This solution creatively uses an improved quantum convolutional long-short-term neural network that combines risk framework analysis and random forest to optimize preoperative care plans. By constructing a risk factor model, a weighted calculation of care risks is performed. The efficient feature extraction capability of quantum convolution and the time series data processing characteristics of long-short-term memory networks are combined. Finally, the features of the three are integrated and a random forest model is constructed to achieve effective multi-source and multi-feature data processing, thereby providing accurate and comprehensive support for preoperative care plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A schematic diagram of the structure of an intelligent risk assessment system for neurosurgery provided by the present invention;

[0062] Figure 2 A flowchart of the steps performed by the data optimization module;

[0063] Figure 3 A flowchart of the steps performed by the surgical planning module;

[0064] Figure 4 Flowchart of the steps performed for the preoperative care plan optimization module.

[0065] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0066] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0067] Example 1, see Figure 1 , the present invention provides a neurosurgery operation risk intelligent assessment system, including a preoperative data collection module, a data optimization module, a surgery planning design module, a preoperative nursing program optimization module and a preoperative risk comprehensive prediction module;

[0068] The preoperative data collection module is used for preoperative data collection, obtains original data of neurosurgery risk assessment through preoperative data collection, and sends the original data of neurosurgery risk assessment to the data optimization module;

[0069] The data optimization module is used for data optimization, and obtains optimized surgical risk assessment data through data optimization, and sends the optimized surgical risk assessment data to the surgical planning and design module, the preoperative nursing program optimization module and the preoperative risk comprehensive prediction module;

[0070] The surgical planning and design module is used for surgical planning and design, obtains surgical planning and design reference data through surgical planning and design, and sends the surgical planning and design reference data to the preoperative nursing program optimization module;

[0071] The preoperative care program optimization module is used to optimize the preoperative care program, obtain preoperative care program reference data through the preoperative care program optimization, and send the preoperative care program reference data to the preoperative risk comprehensive prediction module;

[0072] The preoperative risk comprehensive prediction module is used for preoperative risk comprehensive prediction. Through preoperative risk comprehensive prediction, reference data for neurosurgery risk assessment is obtained, and intelligent neurosurgery risk assessment is performed.

[0073] By performing the above operations, in the existing intelligent assessment system for neurosurgery risk, there is an existing intelligent assessment method or system for neurosurgery risk that mainly relies on the combined perspective of preoperative and postoperative assessment and prediction. However, since neurosurgery involves more complex nursing conditions before surgery and patients also have a variety of preoperative nursing risks, the existing methods for comprehensive preoperative surgical risk assessment of neurosurgery are not detailed enough and the referenceability is not high enough, and the correlation between preoperative analysis and the implementation of the operation itself is not enough. Technical problems, this solution creatively adopts the combination of surgical planning design and preoperative nursing plan optimization to conduct intelligent assessment of neurosurgery risk. By predicting risks and designing plans for the surgical planning design itself, and also conducting a joint analysis of preoperative nursing for neurosurgery, it can effectively combine the specific needs of the operation for nursing, and can also analyze the implementation of the operation according to the specific situation of the patient, thereby improving the correlation between preoperative analysis and surgical implementation, and also realizing the analysis and evaluation of complex patient data.

[0074] Embodiment 2, this embodiment is based on the above embodiment, see Figure 1 , the preoperative data collection is used to collect multimodal data of neurosurgery patients, specifically, to obtain original data of neurosurgery risk assessment from the medical system through data collection;

[0075] The original data of neurosurgery risk assessment specifically includes basic patient information data, clinical examination data, physiological characteristics data, neurological function examination data, patient symptom data and preoperative assessment data;

[0076] The patient's basic information data specifically includes basic personal information and medical history information; the basic personal information includes name, gender, age, height and weight; the medical history information includes disease history, allergy history, family medical history and surgical history information;

[0077] The clinical examination data specifically refers to vital signs data and laboratory examination data; the vital signs data specifically includes heart rate, respiratory rate, blood pressure, body temperature and blood oxygen saturation data; the laboratory examination data specifically includes blood routine, blood sugar, liver and kidney function, coagulation function, blood lipids and inflammatory index data;

[0078] The physiological characteristic data specifically include electrocardiogram data, vital capacity data, urine routine data and functional imaging characteristic data; the functional imaging characteristic data specifically include magnetic resonance imaging characteristics and cerebral angiography characteristic data;

[0079] The neurological function examination data specifically includes neurological sign data and neurological imaging assessment data; the neurological sign data specifically includes motor examination data, sensory examination data, reflex examination data and cognitive ability examination data; the neurological imaging assessment data specifically refers to electroencephalogram assessment data;

[0080] The patient symptom data specifically includes patient-reported symptom data and symptom quantitative assessment data;

[0081] The preoperative evaluation data specifically include comorbidity screening data, anesthesia risk grading assessment data and body mass index assessment data.

[0082] Embodiment 3, this embodiment is based on the above embodiment, see Figure 2 The data optimization is used to clean, fuse and optimize the original data, specifically to optimize the original data of the neurosurgery risk assessment to obtain optimized surgical risk assessment data, and specifically includes the following steps: data cleaning optimization, data standardization optimization, data fusion optimization, feature optimization and data integration optimization;

[0083] The data cleaning and optimization specifically includes performing data deduplication, missing value processing and outlier processing operations on the original data of the neurosurgery risk assessment to obtain cleaned and optimized data; the missing value processing specifically includes performing missing value processing by using mean filling; the outlier processing specifically includes performing anomaly detection on the physiological characteristic data and performing outlier processing by using the box plot method;

[0084] The data standardization optimization specifically includes performing a data standardization operation on the data in the cleaning optimization data, specifically using Z-score standardization to obtain standardized data;

[0085] The data fusion optimization specifically involves extracting features from artificial image data and performing physiological data correlation analysis to obtain comprehensive patient health data, and performing data fusion optimization by combining multi-source and multi-modal data, and obtaining fused and optimized data by aligning time series data;

[0086] The feature optimization is specifically to perform statistical feature selection and principal component analysis feature dimension reduction operations on the feature data in the fused optimization data to obtain feature optimized data;

[0087] The data integration optimization specifically includes performing data consistency check and time series alignment on the feature optimization data and the fusion optimization data to obtain optimized surgical risk assessment data;

[0088] The optimized surgical risk assessment data specifically includes basic patient information characteristics, comprehensive patient health characteristics and preoperative assessment characteristics;

[0089] The patient's basic information characteristics and comprehensive patient health characteristics are used for surgical planning and design;

[0090] The comprehensive patient health characteristics and preoperative assessment characteristics are used to optimize the preoperative care plan.

[0091] Embodiment 4, this embodiment is based on the above embodiment, see Figure 3 The surgical planning design is used to perform personalized surgical path design, instrument recommendation and surgical difficulty assessment according to the patient's specific symptoms. Specifically, based on the optimized surgical risk assessment data, a differential optimization reinforcement learning method with integrated improved convolution feature extraction is used to perform surgical planning design to obtain surgical planning design reference data, which specifically includes the following steps: constructing a heterogeneous fusion integrated improved convolution layer, constructing a reinforcement learning surgical path planning layer, constructing differential evolution optimized surgical instrument planning, surgical planning design model training and surgical planning design;

[0092] The heterogeneous fusion integrated improved convolution layer is constructed to extract surgical planning features, specifically, a standard convolutional neural network and a pre-trained heterogeneous integrated convolutional image feature extraction model are constructed, multi-source heterogeneous data feature extraction is performed through classification feature extraction, and fusion surgical planning feature data is obtained through feature fusion;

[0093] The standard convolutional neural network is used to extract and optimize numerical data features, specifically referring to a standard convolutional neural network including a convolutional layer, a pooling layer and a fully connected layer, and specifically extracts surgical planning numerical data optimization features from the patient's basic information features;

[0094] The pre-trained heterogeneous integrated convolutional image feature extraction model is used to extract and optimize image data features, specifically referring to the integration of a pre-trained ResNet model, a pre-trained Inception model, and a pre-trained VGG model, specifically extracting surgical planning impact data optimization features from the comprehensive patient health features;

[0095] The pre-trained ResNet model specifically adopts the ResNet50 model;

[0096] The pre-trained Inception model specifically adopts the Inceptionv4 model;

[0097] The pre-trained VGG model specifically adopts the VGG19 model;

[0098] The feature fusion specifically refers to fusing the optimization features of the surgical planning numerical data and the optimization features of the surgical planning impact data to obtain fused surgical planning feature data;

[0099] The reinforcement learning surgical path planning layer is constructed to optimize the design of the specific surgical implementation plan, specifically, based on the fused surgical planning feature data, a reinforcement learning model is constructed to optimize the surgical planning path decision, and the reinforcement learning surgical planning output is obtained through reinforcement learning environment construction and deep Q value learning;

[0100] The reinforcement learning environment construction is specifically to design a parameter environment in combination with the surgical planning design, which specifically includes the definition of the parameter environment, the definition of the state parameters, the definition of the action parameters and the definition of the reward function;

[0101] The parameter environment definition specifically refers to simulating the process environment of neurosurgery, including state parameters, action parameters and reward functions;

[0102] The state parameter definition specifically defines the surgical stages and elements during the surgical process, including imaging characteristics, patient physiological characteristics, and surgical step characteristics;

[0103] The action parameter definitions specifically include tumor removal actions, precision cutting actions, hemostasis operation actions, neurovascular repair suturing actions, tissue repair actions, and intraoperative monitoring actions;

[0104] The reward function definition specifically refers to defining it in combination with the surgical success rate, safety, surgical time and surgical risk, and constructing a reward function to perform iterative deep Q-value learning to obtain a reinforcement learning surgical planning output;

[0105] The calculation formula of the reward function is:

[0106] ;

[0107] In the formula, R t is the reward function, w1 is the weight of the surgical success rate, Success(·) is the surgical success rate function, A t is the action parameter, where t is the iteration index, w2 is the surgical safety weight, Safety(·) is the surgical safety function, S t is the state parameter, w3 is the surgery time weight, Time(·) is the surgery success rate function, w4 is the surgery risk weight, Risk(·) is the surgery risk function;

[0108] The iterative deep Q-value learning specifically adopts a Q-value update algorithm combined with policy gradient optimization to perform reinforcement learning optimization;

[0109] The reinforcement learning surgery planning output specifically includes optimal reference steps for surgery, optimal surgery operation selection, reference surgery risk, and surgery time estimation;

[0110] The construction of differential evolution optimization surgical instrument planning is used to optimize the selection of surgical instruments in surgical planning, specifically to construct a standard differential evolution algorithm, initialize a candidate surgical instrument combination, and perform mutation and crossover operations to optimize the surgical planning instrument selection, thereby obtaining surgical planning instrument selection reference data;

[0111] The surgical planning design model training is specifically to train the surgical planning design model through the heterogeneous fusion integrated improved convolution layer, the reinforcement learning surgical path planning layer and the differential evolution optimized surgical instrument planning, and to evaluate the surgical difficulty by combining the surgical path planning and the surgical instrument planning to obtain the surgical planning design model Model OG ;

[0112] The surgical planning design is specifically to use the surgical planning design model Model OG , performing surgical planning and design according to the optimized surgical risk assessment data, and obtaining surgical planning and design reference data;

[0113] The surgical planning and design reference data specifically includes surgical path reference data, surgical instrument reference data and surgical difficulty assessment data;

[0114] The surgical path reference data specifically includes optimal surgical reference steps, optimal surgical operation selection, surgical reference risks and surgical time estimation;

[0115] The surgical difficulty assessment data specifically includes difficulty quantitative scoring valuation data and difficulty grading data.

[0116] By performing the above operations, in view of the fact that in the existing neurosurgery planning and design methods, the preoperative data involved in neurosurgery has the characteristics of multi-source and multi-modality, and the complex heterogeneous data increases the difficulty of constructing the machine model of the surgical steps, and the existing methods are difficult to effectively connect the multi-source heterogeneous preoperative data to the multi-objective surgical planning and design. The technical problem, this solution creatively adopts the differential optimization reinforcement learning method of integrated improved convolutional feature extraction to carry out surgical planning and design. Through the integration and improvement of various convolutional networks, the processing capability of multi-source and multi-structure data is optimized. At the same time, the differential algorithm and reinforcement learning are combined to design and predict surgical instruments and surgical steps, which simultaneously improves the preoperative foresight and knowability of neurosurgery, and provides a more understandable and usable intelligent solution for the concrete decomposition of surgical risks.

[0117] Embodiment 5, this embodiment is based on the above embodiment, see Figure 4The preoperative care plan optimization is used to generate a preoperative care plan for patients suitable for surgery. Specifically, based on the optimized surgical risk assessment data and the surgical planning design reference data, an improved quantum convolutional long-short-term neural network combining risk framework analysis and random forest is used to optimize the preoperative care plan to obtain preoperative care plan reference data, which specifically includes the following steps: constructing a preoperative care risk factor framework model, constructing a quantum convolutional neural network, constructing a long-short-term memory neural network, integrating feature outputs, constructing a random forest decision subnet, training a preoperative care plan optimization model, and optimizing the preoperative care plan;

[0118] The construction of the preoperative nursing risk factor framework model specifically introduces a nursing risk factor framework model composed of preoperative assessment characteristics and surgical planning design reference data, and performs optimization prediction of the preoperative nursing plan based on the nursing risk factor framework model to obtain preoperative nursing risk assessment data. The calculation formula of the nursing risk factor framework model is:

[0119] ;

[0120] In the formula, Risk AT is the preoperative nursing risk assessment data, Pred(·) is the nursing risk factor prediction function, which is specifically predicted by a linear weighted model, PreCare is the preoperative assessment feature, which is used to represent the preoperative assessment feature in the optimized surgical risk assessment data, PlanDesgin is the surgical planning design feature, which is used to represent the surgical difficulty assessment data in the surgical planning design reference data, Healthy is the patient health condition feature, which is used to represent the comprehensive patient health feature in the optimized surgical risk assessment data;

[0121] The construction of the quantum convolutional neural network is specifically to construct a standard quantum convolutional neural network including a quantum convolutional layer and a quantum pooling layer, and to extract the features of the preoperative care plan and predict the drug management in combination with the optimized surgical risk assessment data, so as to obtain the preoperative care drug management features, and the calculation formula is:

[0122] Med QC =QConv(I,W);

[0123] In the formula, Med QC is the quantum convolution feature output, used to represent the preoperative care drug management feature, QConv(I,W) is the quantum convolution calculation function, I is the input data, used to represent the optimized surgical risk assessment data, and W is the quantum convolution weight;

[0124] The constructing of the long short-term memory neural network specifically constructs a standard long short-term memory neural network including an input gate, a forget gate and an output gate, which is used to extract features of the changing trends of the patient health indicators in the optimized surgical risk assessment data, and to dynamically predict the nursing plan to obtain trend nursing plan prediction data;

[0125] The feature output integration is specifically to perform weighted integration fusion on the output features of the preoperative care risk factor framework model, quantum convolutional neural network and long short-term memory neural network to obtain the optimized feature data of the comprehensive preoperative care plan;

[0126] The construction of the random forest decision subnet specifically includes taking the optimized feature data of the comprehensive preoperative nursing plan as the input of the random forest ensemble learning model, and constructing a standard random forest model for comprehensive optimization of the nursing plan to obtain an optimized output of the nursing plan;

[0127] The preoperative care program optimization model training is specifically to perform preoperative care program optimization model training through the preoperative care risk factor framework model, the quantum convolutional neural network, the long short-term memory neural network, the feature output integration and the random forest decision subnet to obtain the preoperative care program optimization model Model HQLR ;

[0128] The preoperative care program optimization is specifically to use the preoperative care program optimization model Model HQLR , optimizing the preoperative care plan based on the optimized surgical risk assessment data and the surgical planning design reference data to obtain preoperative care plan reference data;

[0129] The preoperative care plan reference data specifically includes a preoperative drug management reference plan, a preoperative fasting arrangement reference plan, a preoperative health status trend monitoring reference plan and preoperative care risk assessment data.

[0130] By performing the above operations, in the optimization of existing neurosurgery preoperative care plans, there is a technical problem that preoperative care combined with surgical planning and design involves relatively complex time trend changes in patients' health conditions and micro-lesion characteristics in medical data, and existing methods are difficult to process and analyze these complex characteristics in a fine-grained manner. This solution creatively uses an improved quantum convolutional long-short-term neural network that combines risk framework analysis and random forests to optimize preoperative care plans. By constructing a risk factor model, a weighted calculation of nursing risks is performed, and the efficient feature extraction capability of quantum convolution and the time series data processing characteristics of long-short-term memory networks are combined. Finally, the features of the three are integrated and a random forest model is constructed to achieve effective multi-source and multi-feature data processing, thereby providing accurate and comprehensive support for preoperative care plans.

[0131] Embodiment 6, this embodiment is based on the above embodiment, see Figure 1 The comprehensive preoperative risk prediction is used to conduct a comprehensive assessment of surgical risks, specifically, based on the optimized surgical risk assessment data, the surgical planning design reference data and the preoperative care program reference data, an integrated learning deep neural network method is used to conduct a comprehensive preoperative risk prediction to obtain reference data for neurosurgery risk assessment;

[0132] The neurosurgery risk assessment reference data specifically includes a neurosurgery planning and design assessment reference plan and a neurosurgery preoperative care plan assessment reference plan.

[0133] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process and method including a series of elements include not only those elements, but also other elements not explicitly listed, or also include elements inherent to such process and method.

[0134] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0135] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A neurosurgery risk intelligent assessment system, characterized by: It includes preoperative data collection module, data optimization module, surgical planning and design module, preoperative nursing program optimization module and preoperative risk comprehensive prediction module; The preoperative data collection module is used for preoperative data collection, obtains original data of neurosurgery risk assessment through preoperative data collection, and sends the original data of neurosurgery risk assessment to the data optimization module; The data optimization module is used for data optimization. Through data optimization, optimized surgical risk assessment data is obtained, and the optimized surgical risk assessment data is sent to the surgical planning and design module, the preoperative care program optimization module and the preoperative risk comprehensive prediction module; the optimized surgical risk assessment data specifically includes basic patient information characteristics, comprehensive patient health characteristics and preoperative assessment characteristics; The surgical planning and design module is used for surgical planning and design, obtains surgical planning and design reference data through surgical planning and design, and sends the surgical planning and design reference data to the preoperative nursing program optimization module; The surgical planning and design specifically adopts a differential optimization reinforcement learning method of integrated improved convolution feature extraction to perform surgical planning and design, and specifically includes the following steps: constructing a heterogeneous fusion integrated improved convolution layer, constructing a reinforcement learning surgical path planning layer, constructing differential evolution optimization surgical instrument planning, surgical planning design model training and surgical planning and design; The preoperative care program optimization module is used to optimize the preoperative care program, obtain preoperative care program reference data through the preoperative care program optimization, and send the preoperative care program reference data to the preoperative risk comprehensive prediction module; The optimization of the preoperative care plan specifically adopts an improved quantum convolutional long-short-term neural network that combines risk framework analysis and random forest to optimize the preoperative care plan, and specifically includes the following steps: constructing a preoperative care risk factor framework model, constructing a quantum convolutional neural network, constructing a long-short-term memory neural network, integrating feature outputs, constructing a random forest decision subnet, training a preoperative care plan optimization model, and optimizing the preoperative care plan; The preoperative risk comprehensive prediction module is used for preoperative risk comprehensive prediction. Through preoperative risk comprehensive prediction, reference data for neurosurgery risk assessment is obtained, and intelligent neurosurgery risk assessment is performed.

2. The intelligent neurosurgery risk assessment system according to claim 1, characterized in that: The preoperative data collection is used to collect multimodal data of neurosurgery patients, specifically to obtain original data for neurosurgery risk assessment from the medical system through data collection; The original data for neurosurgery risk assessment specifically includes patient basic information data, clinical examination data, physiological characteristic data, neurological function examination data, patient symptom data and preoperative assessment data.

3. The intelligent neurosurgery risk assessment system according to claim 2, characterized in that: The data optimization is used to clean, fuse and optimize the original data, specifically to optimize the original data of the neurosurgery risk assessment to obtain optimized surgical risk assessment data, and specifically includes the following steps: data cleaning optimization, data standardization optimization, data fusion optimization, feature optimization and data integration optimization; The optimized surgical risk assessment data specifically includes basic patient information characteristics, comprehensive patient health characteristics and preoperative assessment characteristics; The patient's basic information characteristics and comprehensive patient health characteristics are used for surgical planning and design; The comprehensive patient health characteristics and preoperative assessment characteristics are used to optimize the preoperative care plan.

4. The intelligent neurosurgery risk assessment system according to claim 3, characterized in that: The surgical planning design is used to perform personalized surgical path design, instrument recommendation and surgical difficulty assessment according to the patient's specific symptoms. Specifically, based on the optimized surgical risk assessment data, a differential optimization reinforcement learning method of integrated improved convolution feature extraction is used to perform surgical planning design to obtain surgical planning design reference data, and specifically includes the following steps: constructing a heterogeneous fusion integrated improved convolution layer, constructing a reinforcement learning surgical path planning layer, constructing differential evolution optimized surgical instrument planning, surgical planning design model training and surgical planning design; The heterogeneous fusion integrated improved convolution layer is constructed to extract surgical planning features, specifically, a standard convolutional neural network and a pre-trained heterogeneous integrated convolutional image feature extraction model are constructed, multi-source heterogeneous data feature extraction is performed through classification feature extraction, and fusion surgical planning feature data is obtained through feature fusion; The feature fusion specifically refers to fusing the optimization features of the surgical planning numerical data and the optimization features of the surgical planning impact data to obtain fused surgical planning feature data; The reinforcement learning surgical path planning layer is constructed to optimize the design of the specific surgical implementation plan, specifically, based on the fused surgical planning feature data, a reinforcement learning model is constructed to optimize the surgical planning path decision, and the reinforcement learning surgical planning output is obtained through reinforcement learning environment construction and deep Q value learning; The reinforcement learning environment construction is specifically to design a parameter environment in combination with the surgical planning design, which specifically includes the definition of the parameter environment, the definition of the state parameters, the definition of the action parameters and the definition of the reward function; The parameter environment definition specifically refers to simulating the process environment of neurosurgery, including state parameters, action parameters and reward functions; The state parameter definition specifically defines the surgical stages and elements during the surgical process, including imaging characteristics, patient physiological characteristics, and surgical step characteristics; The action parameter definitions specifically include tumor removal actions, precision cutting actions, hemostasis operation actions, neurovascular repair suturing actions, tissue repair actions, and intraoperative monitoring actions; The reward function definition specifically refers to defining it in combination with the surgical success rate, safety, surgical time and surgical risk, and constructing a reward function to perform iterative deep Q-value learning to obtain a reinforcement learning surgical planning output; The calculation formula of the reward function is: ； In the formula, R t is the reward function, w1 is the weight of the surgical success rate, Success(·) is the surgical success rate function, A t is the action parameter, where t is the iteration index, w2 is the surgical safety weight, Safety(·) is the surgical safety function, S t is the state parameter, w3 is the surgery time weight, Time(·) is the surgery success rate function, w4 is the surgery risk weight, Risk(·) is the surgery risk function; The iterative deep Q-value learning specifically adopts a Q-value update algorithm combined with policy gradient optimization to perform reinforcement learning optimization; The reinforcement learning surgery planning output specifically includes optimal reference steps for surgery, optimal surgery operation selection, reference surgery risk, and surgery time estimation; The construction of differential evolution optimization surgical instrument planning is used to optimize the selection of surgical instruments in surgical planning, specifically to construct a standard differential evolution algorithm, initialize a candidate surgical instrument combination, and perform mutation and crossover operations to optimize the surgical planning instrument selection, thereby obtaining surgical planning instrument selection reference data; The surgical planning design model training is specifically to train the surgical planning design model through the heterogeneous fusion integrated improved convolution layer, the reinforcement learning surgical path planning layer and the differential evolution optimized surgical instrument planning, and to evaluate the surgical difficulty by combining the surgical path planning and the surgical instrument planning to obtain the surgical planning design model Model OG ; The surgical planning design is specifically to use the surgical planning design model Model OG , based on the optimized surgical risk assessment data, surgical planning and design are carried out to obtain surgical planning and design reference data.

5. The intelligent neurosurgery risk assessment system according to claim 4, characterized in that: The surgical planning and design reference data specifically includes surgical path reference data, surgical instrument reference data and surgical difficulty assessment data; The surgical path reference data specifically includes optimal surgical reference steps, optimal surgical operation selection, surgical reference risks and surgical time estimation; The surgical difficulty assessment data specifically includes difficulty quantitative scoring valuation data and difficulty grading data.

6. The intelligent neurosurgery risk assessment system according to claim 5, characterized in that: The preoperative care plan optimization is used to generate a preoperative care plan for patients suitable for surgery. Specifically, based on the optimized surgical risk assessment data and the surgical planning design reference data, an improved quantum convolutional long-short term neural network combining risk framework analysis and random forest is used to optimize the preoperative care plan to obtain preoperative care plan reference data. Specifically, the following steps are included: constructing a preoperative care risk factor framework model, constructing a quantum convolutional neural network, constructing a long-short term memory neural network, integrating feature outputs, constructing a random forest decision subnet, training a preoperative care plan optimization model, and optimizing the preoperative care plan; The constructing of the preoperative nursing risk factor framework model specifically includes introducing a nursing risk factor framework model composed of preoperative assessment characteristics and surgical planning design reference data, and performing optimization prediction of the preoperative nursing plan based on the nursing risk factor framework model to obtain preoperative nursing risk assessment data; The constructing of the quantum convolutional neural network specifically constructs a standard quantum convolutional neural network including a quantum convolutional layer and a quantum pooling layer, and extracts preoperative care plan features and predicts drug management in combination with the optimized surgical risk assessment data to obtain preoperative care drug management features; The constructing of the long short-term memory neural network specifically constructs a standard long short-term memory neural network including an input gate, a forget gate and an output gate, which is used to extract features of the changing trends of the patient health indicators in the optimized surgical risk assessment data, and to dynamically predict the nursing plan to obtain trend nursing plan prediction data; The feature output integration is specifically to perform weighted integration fusion on the output features of the preoperative care risk factor framework model, quantum convolutional neural network and long short-term memory neural network to obtain the optimized feature data of the comprehensive preoperative care plan; The construction of the random forest decision subnet specifically includes taking the optimized feature data of the comprehensive preoperative nursing plan as the input of the random forest ensemble learning model, and constructing a standard random forest model for comprehensive optimization of the nursing plan to obtain an optimized output of the nursing plan; The preoperative care program optimization model training is specifically to perform preoperative care program optimization model training through the preoperative care risk factor framework model, the quantum convolutional neural network, the long short-term memory neural network, the feature output integration and the random forest decision subnet to obtain the preoperative care program optimization model Model HQLR ; The preoperative care program optimization is specifically to use the preoperative care program optimization model Model HQLR , based on the optimized surgical risk assessment data and the surgical planning design reference data, the preoperative care plan is optimized to obtain preoperative care plan reference data.

7. The intelligent neurosurgery risk assessment system according to claim 6, characterized in that: The preoperative care plan reference data specifically includes a preoperative drug management reference plan, a preoperative fasting arrangement reference plan, a preoperative health status trend monitoring reference plan and preoperative care risk assessment data.

8. The intelligent neurosurgery risk assessment system according to claim 7, characterized in that: The comprehensive preoperative risk prediction is used to perform a comprehensive assessment of surgical risks, specifically, based on the optimized surgical risk assessment data, the surgical planning design reference data and the preoperative care program reference data, an integrated learning deep neural network method is used to perform a comprehensive preoperative risk prediction to obtain reference data for neurosurgery risk assessment; The neurosurgery risk assessment reference data specifically includes a neurosurgery planning and design assessment reference plan and a neurosurgery preoperative care plan assessment reference plan.

Images