Neurosurgical Vascular Care Monitoring System Based on Medical Information Processing

Abstract

The present invention discloses a neurosurgery vascular care monitoring system based on medical information processing, including a medical information collection module, a medical data optimization module, a neurosurgery vascular care evaluation model establishment module and a neurosurgery vascular care monitoring module. The present invention belongs to the field of medical information processing, and specifically refers to a neurosurgery vascular care monitoring system based on medical information processing. The scheme excludes pseudo-anomalies by scoring abnormal path lengths, increases the sensitivity to mild abnormalities and severe abnormal care states based on a category-weighted marginal contribution mechanism, optimizes historical medical care data through abnormality detection, screening and feature selection, and improves the nursing monitoring effect; based on efficient feature extraction of spatiotemporal dependency, the spatial interaction relationship between dynamic physiological signals is captured, and the short-term fluctuations and long-term trends of physiological signals are captured at the same time. According to the relationship differences between physiological signals, dynamic attention weights are introduced to improve the nursing monitoring effect.

Description

Technical Field

[0001] The present invention relates to the field of medical information processing, and specifically refers to a neurosurgical vascular care monitoring system based on medical information processing. Background Art

[0002] A neurosurgical vascular care monitoring system is a system that combines advanced medical equipment and data analysis, focusing on the real-time monitoring of the cerebrovascular status and nursing assessment of neurosurgical patients. It integrates multimodal data and intelligent analysis techniques to provide medical staff with accurate condition assessments and personalized nursing recommendations, helping to improve patient prognosis and optimize the allocation of nursing resources. However, general neurosurgical vascular care monitoring systems have problems such as a lack of dynamic anomaly detection and insufficient feature screening relying on abnormal signals, which leads to low accuracy in nursing status assessment and poor nursing monitoring effects; general neurosurgical vascular care monitoring systems have problems such as poor ability to capture dynamic physiological signals and insufficient collaborative analysis ability, resulting in inaccurate nursing status assessment, and being less sensitive to changes in the distribution of nursing data, leading to poor nursing monitoring effects. Summary of the Invention

[0003] In view of the above situation, to overcome the defects of the prior art, the present invention provides a neurosurgical vascular care monitoring system based on medical information processing. Aiming at the problems of general neurosurgical vascular care monitoring systems, such as a lack of dynamic anomaly detection and insufficient feature screening relying on abnormal signals, which leads to low accuracy in nursing status assessment and poor nursing monitoring effects, this solution uses a dynamic threshold mechanism with a sliding window and an abnormal path length score to exclude pseudo anomalies caused by sensor failures and data fluctuations, increases the sensitivity to mild and severe abnormal nursing states based on a category-weighted marginal contribution mechanism, introduces a combined index of mutual information and correlation redundancy regularization for feature selection, optimizes historical medical nursing data through anomaly detection, screening, and feature selection, and thus improves the accuracy of subsequent nursing status assessment and the nursing monitoring effect; aiming at the problems of general neurosurgical vascular care monitoring systems, such as poor ability to capture dynamic physiological signals and insufficient collaborative analysis ability, resulting in inaccurate nursing status assessment, and being less sensitive to changes in the distribution of nursing data, leading to poor nursing monitoring effects, this solution is based on spatio-temporal dependence for efficient feature extraction to capture the spatial interaction relationships between dynamic physiological signals, and at the same time captures the short-term fluctuations and long-term trends of physiological signals to improve the sensitivity to the trend of disease changes. In view of the relationship differences between physiological signals, a multi-head attention mechanism based on dynamic attention weights is introduced to improve the prediction accuracy of nursing status, and thus improve the nursing monitoring effect.

[0004] The technical solution adopted by the present invention is as follows: The neurosurgical vascular care monitoring system based on medical information processing provided by the present invention includes a medical information collection module, a medical data optimization module, a neurosurgical vascular care assessment model establishment module, and a neurosurgical vascular care monitoring module;

[0005] The medical information collection module collects historical medical care data;

[0006] The medical data optimization module constructs data input for care status assessment through dynamic anomaly detection, feature screening, and optimal feature selection;

[0007] The neurosurgical vascular care assessment model establishment module extracts and fuses time-dependent and spatial interaction features through a residual network, a convolutional gated recurrent unit, and a multi-head attention mechanism, and finally realizes care status prediction;

[0008] The neurosurgical vascular care monitoring module realizes status assessment for real-time collected data based on the established neurosurgical vascular care assessment model, and further realizes care monitoring.

[0009] Furthermore, in the medical information collection module, the historical medical care data includes dynamic physiological signals, chemical metabolism data, time, and care status; the care status includes normal, mildly abnormal, and severely abnormal; the care status is used as a data label.

[0010] Furthermore, the medical data optimization module specifically includes the following:

[0011] Initial anomaly detection; for dynamic physiological signal data, detect outliers caused by sensor failures, and introduce a dynamic threshold mechanism based on a sliding window, expressed as: ; ; ; construct historical sequence features for the historical medical care data after initial anomaly removal; where, IF(·) is the anomaly detection score; x is the data point; mT is the number of time steps, t is the time step index; E(·) is the anomaly path length; n is the number of samples; c is the normalization coefficient; is the dynamic anomaly weight; is the data point value at time step t; k is the neighborhood size, ik is the neighbor index; d(·) is the Euclidean distance; is the nearest neighbor point; is the maximum nearest neighbor distance; is the sliding window the median value within; and are the standard deviation and median absolute deviation respectively;

[0012] Preliminary feature screening; in medical data, different features contribute differently to cerebrovascular abnormalities. Therefore, quantify the marginal contribution of each feature and introduce category-weighted marginal contribution; thus, perform preliminary feature screening, expressed as: ; ; ; where is the contribution value of the i1-th feature; is the size of the feature subset S; is the size of the full feature set N; is the category weight; and are the weight increases for abnormal states; f1, f2, and f3 are the normal state, mild abnormal state, and severe abnormal state, respectively; is the performance metric, obtained based on a pre-trained regression model; is the sample weight; i1 is the feature index; i is the data point index; is the number of data points included in the category; is the category to which the data point belongs;

[0013] Optimal feature selection; for the relevant features obtained from preliminary feature screening, further select the optimal feature subset, expressed as: ; ; ; ; ; where is the balance weight; bZ is the final result of feature selection; m is the number of selected features; and are the feature weights; is the current feature subset; is the true care status label of the i-th sample; is the predicted label of the i-th sample based on the feature subset; f is a single feature; c is the target care status; I(·) is the mutual information; f i1 and f j1 are two features in the feature subset; Var(·) is the variance; y and are the true label and predicted label, respectively; is the KL divergence; is the normalization constant; is the balance factor.

[0014] Furthermore, the neurosurgical vascular care assessment model establishment module is based on the historical medical care data processed by the medical data optimization module, specifically including the following:

[0015] Feature extraction; Physiological signals and chemical data have obvious time dependence. Abnormal short-term fluctuations can be precursors of impaired cerebrovascular regulation function, while long-term trend changes reflect the deterioration of the condition. Therefore, the spatial correlation between physiological signals is extracted through a residual network to capture the interaction relationship between different monitoring signals, expressed as: ; The dynamic time features are extracted through a convolutional gated recurrent unit and based on convolutional operations to capture the time dependence of physiological signals, expressed as: ; ; Among them, is the historical medical care data input at the t-i time step; W a is the input weight matrix; is the bias vector; Softmax(·) is the Softmax function; is the output of the residual network; is the spatial correlation feature, reflecting the interaction relationship between different physiological signals; is the update gate; and are the convolutional weight matrix and the recurrent weight matrix respectively; is the bias vector of the convolutional gated recurrent unit; is the Sigmoid activation function; is the candidate state at the current time step; is the final hidden state; is the hidden state;

[0016] Feature fusion; The changes of different physiological signals in the time dimension have different importance for the prediction of abnormal cerebral blood flow. The relationship between features is captured through the attention operation in parallel of the multi-head attention mechanism, expressed as: ; ; Global pooling is performed based on adaptive pooling, expressed as: ; The feature regularization term is added, expressed as: ; Finally, the nursing status is predicted through a fully connected layer; The model uses the cross-entropy loss function and optimizes the parameters by the gradient descent method; Among them, MHd(·) is the output of the multi-head attention mechanism; Q, K, and V are the query matrix, the key matrix, and the value matrix respectively; , and are the outputs of the first, the second, and the n2-th attention heads respectively; , and are the corresponding weights; is the linear mapping weight; is the concatenation operation; and are the results of the i2-th linear transformation; and is the result of the j2-th linear transformation; Sim(·) is; is the pooled feature; is the adaptive pooling operation; is the size of the output feature; x is the input feature matrix, including short-term fluctuations and long-term trend features; is the feature regularization term of the fully connected layer; is the weight matrix of the fully connected layer; is the regularization strength; is the L2 norm.

[0017] Furthermore, the neurosurgical vascular care monitoring module is based on the established neurosurgical vascular care assessment model, which collects medical care data in real time and uses the model output as the evaluation result of the collected medical care data. When the evaluation result is severely abnormal, a warning is issued to the nursing staff.

[0018] The beneficial effects of the present invention using the above solution are as follows:

[0019] (1) Aiming at the problems of the general neurosurgical vascular care monitoring system, such as the lack of dynamic anomaly detection, insufficient feature screening for abnormal signals, resulting in low accuracy of nursing status evaluation and poor nursing monitoring effect. This solution uses a dynamic threshold mechanism with a sliding window and an abnormal path length score to exclude pseudo anomalies caused by sensor failures and data fluctuations, increases the sensitivity to mild and severe abnormal nursing status based on the category-weighted marginal contribution mechanism, introduces a combined index of mutual information and correlation redundancy regularization for feature selection, optimizes historical medical care data through anomaly detection, screening, and feature selection, and thus improves the accuracy of subsequent nursing status evaluation and the nursing monitoring effect.

[0020] (2) Aiming at the problems of the general neurosurgical vascular care monitoring system, such as poor ability to capture dynamic physiological signals and insufficient collaborative analysis ability, resulting in inaccurate nursing status evaluation, and lack of sensitivity to changes in the distribution of nursing data, resulting in poor nursing monitoring effect. This solution is based on spatio-temporal dependence for efficient feature extraction, captures the spatial interaction relationship between dynamic physiological signals, and simultaneously captures the short-term fluctuations and long-term trends of physiological signals, improves the sensitivity to the trend of disease changes. Aiming at the relationship differences between physiological signals, a multi-head attention mechanism based on dynamic attention weights is introduced to improve the prediction accuracy of nursing status, and thus improve the nursing monitoring effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a schematic flow chart of the neurosurgical vascular care monitoring system based on medical information processing provided by the present invention;

[0022] Figure 2It is a schematic flowchart of the medical data optimization module.

[0023] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation to the present invention. Detailed implementation manners

[0024] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a 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 those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0025] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention.

[0026] Example 1, refer to Figure 1 The neurosurgical vascular care monitoring system based on medical information processing provided by the present invention includes a medical information acquisition module, a medical data optimization module, a neurosurgical vascular care assessment model establishment module, and a neurosurgical vascular care monitoring module;

[0027] The data acquisition module acquires historical medical care data; and sends the data to the medical data optimization module;

[0028] The medical data optimization module constructs data input for care status assessment through dynamic anomaly detection, feature screening, and optimal feature selection; and sends the data to the neurosurgical vascular care assessment model establishment module;

[0029] The neurosurgical vascular care assessment model establishment module extracts and fuses time-dependent and spatial interaction features through a residual network, a convolutional gated recurrent unit, and a multi-head attention mechanism, and finally realizes care status prediction; and sends the data to the neurosurgical vascular care monitoring module;

[0030] The neurosurgical vascular care monitoring module realizes status assessment of the real-time collected data based on the established neurosurgical vascular care assessment model, and further realizes care monitoring.

[0031] Example 2, refer to Figure 1This embodiment is based on the above embodiment. In the medical information acquisition module, the historical medical care data includes dynamic physiological signals, chemical metabolic data, time and care status; the care status includes normal, mild abnormality and severe abnormality; the care status is used as a data label; the dynamic physiological signals include intracranial pressure, cerebral perfusion pressure, cerebral blood flow velocity, cerebral oxygen saturation, heart rate, blood pressure, cardiac output, patient position and activity status; the chemical metabolic data includes cerebrospinal fluid detection data and blood detection data.

[0032] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the medical data optimization module specifically includes the following contents:

[0033] Preliminary anomaly detection: For dynamic physiological signal data, detect anomalies caused by sensor failures and introduce a dynamic threshold mechanism based on sliding windows, expressed as: ; ; ; Construct historical sequence features for historical medical care data after preliminary anomaly removal; where IF(·) is the anomaly detection score; x is the data point; mT is the number of time steps, t is the time step index; E(·) is the anomaly path length; n is the number of samples; c is the normalization coefficient; is the dynamic anomaly weight; is the data point value at time step t; k is the neighborhood size, ik is the neighbor index; d(·) is the Euclidean distance; is the nearest neighbor point; is the maximum neighbor distance; is a sliding window[ ] is the median value within ; and are the standard deviation and median absolute deviation, respectively;

[0034] Preliminary feature screening: In medical data, different features contribute differently to cerebrovascular abnormalities, so the marginal contribution of each feature is quantified, and the category-weighted marginal contribution is introduced to increase the sensitivity to abnormal states; thus, preliminary feature screening is performed; expressed as: ; ; ;in, is the contribution value of the i1th feature; is the size of the feature subset S; is the size of the feature set N; is the category weight; and They are the weights of abnormal states; f1, f2 and f3 are normal state, mild abnormal state and severe abnormal state respectively; is a performance metric, obtained based on a pre-trained regression model; is the sample weight; i1 is the feature index; i is the data point index; is the number of data points contained in the category; is the category to which the data point belongs;

[0035] Optimal feature selection: for the correlation features of the preliminary feature screening, further select the optimal feature subset to enhance the combined analysis capability of dynamic physiological signals and chemical metabolism data; expressed as: ; ; ; ; ;in, is the balance weight; bZ is the final result of feature selection; m is the number of selected features; and is the feature weight; is the current feature subset; is the true nursing status label of the i-th sample; is the predicted label of the ith sample based on the feature subset; f is a single feature; c is the target care status; I(·) is the mutual information; f i1 and f j1 are two features in the feature subset; Var(·) is the variance; y and They are the true label and the predicted label respectively; is the KL divergence; is the normalization constant; is the balancing factor.

[0036] By performing the above operations, in view of the lack of dynamic anomaly detection and insufficient feature screening of abnormal signal dependence in general neurosurgery vascular care monitoring systems, which leads to low accuracy of nursing status assessment and poor nursing monitoring effect, this scheme uses the dynamic threshold mechanism of the sliding window and the abnormal path length scoring to eliminate pseudo anomalies caused by sensor failures and data fluctuations, increases the sensitivity to mild and severe abnormal nursing states based on the category-weighted marginal contribution mechanism, introduces a combined indicator of mutual information and correlation redundancy regularization for feature selection, and optimizes historical medical care data through anomaly detection, screening and feature selection, thereby improving the accuracy of subsequent nursing status assessment and improving nursing monitoring effects.

[0037] Example 4, see Figure 1 This embodiment is based on the above embodiment. The neurosurgery vascular care evaluation model establishment module is based on the historical medical care data processed by the medical data optimization module, and specifically includes the following contents:

[0038] Feature extraction; physiological signals and chemical data have obvious time dependence. Abnormal short-term fluctuations can be precursors of impaired cerebrovascular regulation function, while long-term trend changes reflect the deterioration of the condition. Therefore, the spatial correlation between physiological signals is extracted through a residual network to capture the interaction relationship between different monitoring signals, expressed as: ; The dynamic time features are extracted through a convolutional gated recurrent unit and based on convolutional operations to capture the time dependence of physiological signals, expressed as: ; ; Among them, is the historical medical care data input at the t-i time step; W a is the input weight matrix; is the bias vector; Softmax(·) is the Softmax function; is the output of the residual network; is the spatial correlation feature, reflecting the interaction relationship between different physiological signals; is the update gate; and are the convolutional weight matrix and the recurrent weight matrix respectively; is the convolutional gated recurrent unit bias vector; is the Sigmoid activation function; is the candidate state at the current time step; is the final hidden state; is the hidden state;

[0039] Feature fusion; the changes of different physiological signals in the time dimension have different importance for predicting abnormal cerebral blood flow. The relationship between features is captured through the attention operation in parallel by the multi-head attention mechanism, expressed as: ; ; Global pooling is performed based on adaptive pooling, expressed as: ; The feature regularization term is added, expressed as: ; Finally, the care status is predicted through a fully connected layer; the model uses the cross-entropy loss function and optimizes the parameters by the gradient descent method; among them, MHd(·) is the output of the multi-head attention mechanism; Q, K, and V are the query matrix, the key matrix, and the value matrix respectively; , and are the outputs of the 1st, 2nd, and n2nd attention heads respectively; , and are the corresponding weights; is the linear mapping weight; is the concatenation operation; and are the results of the i2nd linear transformation; and is the result of the j2-th linear transformation; Sim(·) is; are the pooled features; is the adaptive pooling operation; is the size of the output features; x is the input feature matrix, including short-term fluctuations and long-term trend features; is the feature regularization term of the fully connected layer; is the weight matrix of the fully connected layer; is the regularization strength; is the L2 norm.

[0040] By performing the above operations, for the general neurosurgical vascular care monitoring system, there are problems such as poor ability to capture dynamic physiological signals, insufficient collaborative analysis ability, resulting in inaccurate assessment of the care status, and lack of sensitivity to changes in the distribution of care data, resulting in poor care monitoring effects. This solution is based on efficient feature extraction depending on time and space, captures the spatial interaction relationships between dynamic physiological signals, and at the same time captures the short-term fluctuations and long-term trends of physiological signals, improving the sensitivity to the trend of the condition change. In view of the relationship differences between physiological signals, a multi-head attention mechanism based on dynamic attention weights is introduced to improve the prediction accuracy of the care status, thereby improving the care monitoring effect.

[0041] Example Five, refer to Figure 1 , this example is based on the above example. The neurosurgical vascular care monitoring module is based on the established neurosurgical vascular care assessment model, and medical care data is collected in real time. The model output is used as the assessment result of the collected medical care data. When the assessment result is severely abnormal, a warning is given to the nursing staff.

[0042] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variation thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements but also other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0043] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0044] The above describes the present invention and its implementation manners. Such description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. In general, if those of ordinary skill in the art are inspired by it and, without departing from the gist of the present invention, design similar structural manners and embodiments to this technical solution without creative efforts, they shall fall within the protection scope of the present invention.

Claims

1. Neurosurgery vascular care monitoring system based on medical information processing, characterized by: The system includes a medical information collection module, a medical data optimization module, a neurosurgery vascular care evaluation model establishment module, and a neurosurgery vascular care monitoring module; The medical information collection module collects historical medical care data; The medical data optimization module constructs data input for nursing status assessment through preliminary anomaly detection, preliminary feature screening and optimal feature selection; The neurosurgery vascular care evaluation model establishment module is based on the historical medical care data processed by the medical data optimization module, and extracts and fuses time-dependent and spatial interaction features through residual networks, convolutional gated recurrent units, and multi-head attention mechanisms, and finally constructs a neurosurgery vascular care evaluation model; The neurosurgery vascular care monitoring module implements status evaluation of real-time collected data based on the established neurosurgery vascular care evaluation model, thereby realizing care monitoring; The medical data optimization module includes the following contents: preliminary anomaly detection; for dynamic physiological signal data, detecting anomalies caused by sensor failures, and introducing a dynamic threshold mechanism based on a sliding window, which is expressed as: ; ; ; Construct historical sequence features for historical medical care data after preliminary anomaly removal; where IF(·) is the anomaly detection score; x is the data point; mT is the number of time steps, t is the time step index; E(·) is the anomaly path length; n is the number of samples; c is the normalization coefficient; is the dynamic anomaly weight; is the data point value at time step t; k is the neighborhood size, ik is the neighbor index; d(·) is the Euclidean distance; is the nearest neighbor point; is the maximum neighbor distance; is a sliding window[ ] is the median value within ; and are the standard deviation and median absolute deviation respectively.

2. The neurosurgery vascular care monitoring system based on medical information processing according to claim 1 is characterized in that: The medical data optimization module specifically includes the following contents: Preliminary anomaly detection; Preliminary feature screening; quantify the marginal contribution of each feature and introduce category-weighted marginal contribution; thus preliminary feature screening is performed; expressed as: ; ; ;in, is the contribution value of the i1th feature; is the size of the feature subset S; is the size of the feature set N; is the category weight; and They are the weights of abnormal states; f1, f2 and f3 are normal state, mild abnormal state and severe abnormal state respectively; is a performance metric, obtained based on a pre-trained regression model; is the sample weight; i1 is the feature index; i is the data point index; is the number of data points contained in the category; is the category to which the data point belongs; Optimal feature selection: For the correlation features of the preliminary feature screening, further select the optimal feature subset, which is expressed as: ; ; ; ; ;in, is the balance weight; bZ is the final result of feature selection; m is the number of selected features; and is the feature weight; is the current feature subset; is the true nursing status label of the i-th sample; is the predicted label of the ith sample based on the feature subset; f is a single feature; c is the target care status; I(·) is the mutual information; f i1 and f j1 are two features in the feature subset; Var(·) is the variance; y and They are the true label and the predicted label respectively; is the KL divergence; is the normalization constant; is the balancing factor.

3. The neurosurgery vascular care monitoring system based on medical information processing according to claim 2 is characterized in that: The neurosurgery vascular care evaluation model establishment module is based on the historical medical care data processed by the medical data optimization module and specifically includes the following contents: Feature extraction: The spatial correlation between physiological signals is extracted through the residual network to capture the interactive relationship between different monitoring signals, which is expressed as: ; Dynamic time features are extracted based on the convolution operation through the convolution gated recurrent unit to capture the time dependency of physiological signals, which is expressed as: ; ;in, is the historical medical care data input at time step ti; W a is the input weight matrix; is the bias vector; Softmax(·) is the Softmax function; is the residual network output; It is a spatial correlation feature, reflecting the interactive relationship between different physiological signals; It is the update gate; and They are the convolution weight matrix and the circulation weight matrix respectively; is the convolutional gated recurrent unit bias vector; is the Sigmoid activation function; is the candidate state for the current time step; is the final hidden state; is a hidden state; Feature fusion: Capture the relationship between features through the parallel attention operation of the multi-head attention mechanism, expressed as: ; ; Global pooling based on adaptive pooling is expressed as: ; Add feature regularization term, expressed as: ; Finally, the nursing status is predicted through the fully connected layer; the model uses the cross entropy loss function and the gradient descent method to optimize the parameters; where MHd(·) is the output of the multi-head attention mechanism; Q, K and V are the query matrix, key matrix and value matrix respectively; , and They are the outputs of the 1st, 2nd, and n2th attention heads respectively; , and is the corresponding weight; is the linear mapping weight; It is a splicing operation; and is the i2th linear transformation result; and is the j2th linear transformation result; Sim(·) is; is the feature after pooling; is an adaptive pooling operation; is the size of the output feature; x is the input feature matrix, including short-term fluctuations and long-term trend features; is the feature regularization term of the fully connected layer; is the weight matrix of the fully connected layer; is the regularization strength; is the L2 norm.

4. The neurosurgery vascular care monitoring system based on medical information processing according to claim 3 is characterized in that: In the medical information collection module, the historical medical care data includes dynamic physiological signals, chemical metabolism data, time and care status; the care status includes normal, mild abnormality and severe abnormality; and the care status is used as a data label.

5. The neurosurgery vascular care monitoring system based on medical information processing according to claim 4 is characterized in that: The neurosurgery vascular care monitoring module is based on an established neurosurgery vascular care evaluation model, collects medical care data in real time, and uses the model output as the evaluation result of the collected medical care data. When the evaluation result is seriously abnormal, the nursing staff is warned.

Images