Visual anesthesia monitoring system for perioperative patient

Abstract

The invention discloses a perioperative patient anesthesia visual monitoring system which comprehensively analyzes basic body indexes, test indexes and operation information of a patient object to be anesthetized and ASA grading information given by a doctor by adopting a deep learning-based data processing technology. The method comprises the following steps: firstly, extracting semantic feature representation of each piece of patient information through vectorization processing, then performing context semantic association reinforcement on the semantic feature field effect distribution of each piece of patient information, and performing cross-modal semantic association fusion on each piece of patient information on the basis to realize deep global understanding of the patient information. Therefore, intelligent assessment and visual display of the anesthesia risk of the patient are realized. In this way, more comprehensive and accurate decision support can be provided for anesthetists, and the efficiency and safety of anesthesia management of patients in the perioperative period are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent monitoring, and more specifically, to a visual monitoring system for perioperative anesthesia of patients. Background Art

[0002] With the rapid development of modern medical technology, surgical treatment has become one of the important treatment methods for many diseases. During the operation, the patient's anesthesia management is crucial to ensure the safety of the operation and improve the success rate of the operation. In traditional anesthesia management, it mainly relies on medical staff to directly observe the patient's vital signs and monitor simple physiological parameters such as heart rate, blood pressure, respiratory rate, etc. However, this method is often limited by the experience and judgment of medical staff, and the identification of complex cases or potential risk factors is not accurate and comprehensive enough.

[0003] With the continuous advancement of medical technology, especially the rapid development of information technology, more and more medical devices have begun to introduce intelligent and digital monitoring methods. In the field of anesthesia, although there are some monitoring systems based on physiological parameters, most of these systems can only provide single, isolated physiological data monitoring, lacking a comprehensive assessment and in-depth understanding of the patient's overall condition, resulting in a one-sided and rough assessment of the patient's anesthesia risk.

[0004] Therefore, an optimized visual monitoring system for perioperative anesthesia in patients is desired. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a visual monitoring system for perioperative anesthesia of patients, which comprehensively analyzes the basic physical indicators, laboratory indicators, surgical information and ASA grading information given by doctors of anesthetized patients by using data processing technology based on deep learning. First, the semantic feature representation of each patient information is extracted through vector processing, and then the contextual semantic association is strengthened based on the field effect distribution of the semantic features of each patient information. On this basis, the cross-modal semantic association of each patient information is fused to achieve a deep global understanding of the patient information, and then realize the intelligent assessment and visual display of the patient's anesthesia risk. In this way, more comprehensive and accurate decision support can be provided to anesthesiologists, significantly improving the efficiency and safety of perioperative patient anesthesia management.

[0006] According to one aspect of the present application, a perioperative patient anesthesia visualization monitoring system is provided, comprising:

[0007] A patient object data collection module is used to collect the original data of the patient object to be anesthetized, wherein the original data includes basic physical indicators, laboratory indicators, operation information and ASA classification given by the doctor;

[0008] A vectorization processing module, used for performing vectorization processing on each data item in the original data of the patient object to be anesthetized to obtain a set of encoding feature vectors of the original data of the patient object;

[0009] A patient information context semantic enhancement module, used for performing semantic feature enhancement on the set of patient object original data encoding feature vectors to obtain a set of semantically enhanced patient object original data encoding feature vectors;

[0010] A multi-source data cross-modal fusion module, used for cross-modal fusion of the set of semantically enhanced patient object original data encoding feature vectors to obtain a patient object cross-modal multi-source data fusion feature vector;

[0011] an anesthesia risk probability estimation module, used to determine the anesthesia risk probability based on the cross-modal multi-source data fusion feature vector of the patient object;

[0012] A display module is used to display the anesthesia risk probability on a display screen.

[0013] Compared with the prior art, the present application provides a visual monitoring system for perioperative anesthesia of patients. It uses deep learning-based data processing technology to comprehensively analyze the basic physical indicators, laboratory indicators, surgical information, and ASA grading information given by doctors of anesthetized patients. First, the semantic feature representation of each patient information is extracted through vector processing, and then the contextual semantic association is strengthened based on the field effect distribution of the semantic features of each patient information. On this basis, the cross-modal semantic association of each patient information is fused to achieve a deep global understanding of the patient information, and then the intelligent assessment and visual display of the patient's anesthesia risk can be achieved. In this way, more comprehensive and accurate decision support can be provided to anesthesiologists, significantly improving the efficiency and safety of perioperative patient anesthesia management. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0015] Figure 1 is a block diagram of a perioperative patient anesthesia visualization monitoring system according to an embodiment of the present application;

[0016] Figure 2 A data flow diagram of a visual monitoring system for perioperative anesthesia of patients according to an embodiment of the present application;

[0017] Figure 3 The present invention is a block diagram of a patient information context semantic enhancement module in a perioperative patient anesthesia visualization monitoring system according to an embodiment of the present application. DETAILED DESCRIPTION

[0018] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0019] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0020] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0021] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] With the continuous advancement of medical technology, especially the rapid development of information technology, more and more medical devices have begun to introduce intelligent and digital monitoring methods. In the field of anesthesia, although there are some monitoring systems based on physiological parameters, most of these systems can only provide single, isolated physiological data monitoring, lacking a comprehensive assessment and in-depth understanding of the patient's overall condition, resulting in a one-sided and rough assessment of the patient's anesthesia risk. Therefore, an optimized visual monitoring system for perioperative patient anesthesia is expected.

[0024] In the technical solution of the present application, a visual monitoring system for perioperative anesthesia of patients is proposed. Figure 1 The block diagram is a visual monitoring system for perioperative anesthesia of patients according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a visual monitoring system for perioperative anesthesia of patients according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to an embodiment of the present application, a perioperative patient anesthesia visualization monitoring system 300 includes: a patient object data acquisition module 310, which is used to collect original data of the patient object to be anesthetized, and the original data includes basic physical indicators, laboratory indicators, surgical information and ASA grades given by doctors; a vectorization processing module 320, which is used to vectorize each data item in the original data of the patient object to be anesthetized to obtain a set of patient object original data encoding feature vectors; a patient information context semantic enhancement module 330, which is used to perform semantic feature enhancement on the set of patient object original data encoding feature vectors to obtain a set of semantically enhanced patient object original data encoding feature vectors; a multi-source data cross-modal fusion module 340, which is used to perform cross-modal fusion on the set of semantically enhanced patient object original data encoding feature vectors to obtain a patient object cross-modal multi-source data fusion feature vector; an anesthesia risk probability estimation module 350, which is used to determine the anesthesia risk probability based on the patient object cross-modal multi-source data fusion feature vector; a display module 360, which is used to display the anesthesia risk probability on a display screen.

[0025] In particular, the patient object data acquisition module 310 is used to collect the original data of the patient object to be anesthetized, and the original data includes basic physical indicators, laboratory indicators, surgical information and the ASA grade given by the doctor. It should be understood that the ASA grade is a grading standard formulated by the American Society of Anesthesiologists (ASA) to assess the overall health status of patients before anesthesia. The ASA grade ranges from I to VI, and the higher the level, the worse the patient's health status and the greater the risk of anesthetic complications. Therefore, the ASA grade given by the doctor can preliminarily reveal the patient's overall health status and anesthesia risk level, but because the ASA grading method roughly divides the patient's physical condition into six levels, it is impossible to provide patients with a more targeted risk assessment. Therefore, based on the ASA classification information given by the doctor, this application further combines the patient's basic physical indicators (such as age, weight, height, blood pressure, etc.), laboratory indicators (such as blood routine, liver and kidney function test results, etc.) and surgical information (such as surgery type, estimated duration, surgical site, etc.), and conducts a comprehensive analysis of multi-source data to more accurately assess the patient's anesthesia risk, so that the doctor can intuitively understand the patient's condition and provide more accurate auxiliary guidance for the anesthesiologist's decision-making.

[0026] In particular, the vectorization processing module 320 is used to vectorize each data item in the original data of the patient object to be anesthetized to obtain a set of original data encoding feature vectors of the patient object. Considering that the original data contains different types of information, such as text data (such as test result description, operation description), classification data (such as ASA grading). Therefore, in order to convert each item of information in the original data into a unified numerical vector form, so as to facilitate subsequent comprehensive data analysis, the present application further vectorizes each data item in the original data of the patient object to be anesthetized. In an embodiment of the present application, for the patient's basic physical indicators, test indicators and surgical information, word embedding technology and Transformer model are used to extract contextual semantic features of the three respectively, and generate corresponding numerical vector representations to obtain body mark encoding feature vectors, test indicator encoding feature vectors, and surgical information encoding feature vectors; for the ASA grading information given by the doctor, since it is classified data, the present application uses a unique hot encoding technology to convert it into a binary numerical encoding form to obtain an ASA grading encoding feature vector. Furthermore, the coding features of each data are integrated into a unified feature set to form a set of coded feature vectors of the original data of the patient object, so as to ensure that the subsequent data analysis process can comprehensively consider all relevant information of the patient.

[0027] In particular, the patient information context semantic enhancement module 330 is used to perform semantic feature enhancement on the set of original data encoding feature vectors of the patient objects to obtain a set of semantically enhanced original data encoding feature vectors of the patient objects. Since there is an intrinsic correlation between the patient's basic physical indicators, laboratory indicators, surgical information and the ASA grade given by the doctor, for example, the abnormality of a certain laboratory indicator may be associated with the patient's basic physical indicators and surgical type. Therefore, in order to reveal the interaction and influence between various data items and thus enhance the expressive power of features, the present application proposes a semantic enhancement method based on feature field effect distribution, which analyzes the semantic feature field effect distribution of each item of patient information to learn the interaction effect between features, thereby utilizing its global context association to enhance the deep semantic understanding of each item of patient information and enhance the semantic richness of feature representation. In a specific example of the present application, such as Figure 3As shown, the patient information context semantic enhancement module 330 includes: a feature field effect distribution analysis unit 331, which is used to update the set of patient object original data encoding feature vectors based on the feature field effect distribution of the set of patient object original data encoding feature vectors to obtain a set of updated patient object original data encoding feature vectors; a bidirectional semantic association enhancement unit 332, which is used to perform bidirectional context semantic association enhancement on the set of updated patient object original data encoding feature vectors to obtain the set of semantically enhanced patient object original data encoding feature vectors.

[0028] Specifically, the feature field effect distribution analysis unit 331 is used to update the set of patient object raw data encoding feature vectors based on the feature field effect distribution of the set of patient object raw data encoding feature vectors to obtain a set of updated patient object raw data encoding feature vectors. In an embodiment of the present application, first, a semantic reconstructor is used to perform feature reconstruction on each patient object raw data encoding feature vector in the set of patient object raw data encoding feature vectors to obtain a set of reconstructed patient object raw data encoding feature vectors; that is, a semantic reconstructor is used to perform domain transformation on each patient object raw data encoding feature vector in the set of patient object raw data encoding feature vectors to achieve feature dimension unification and enhance its representation ability in a specific context to generate a set of reconstructed patient object raw data encoding feature vectors. Next, the semantic depth implicit representation value of each reconstructed patient object raw data encoding feature vector in the set of reconstructed patient object raw data encoding feature vectors is calculated to obtain a set of semantic depth implicit representation values ​​of patient object raw data; by calculating the semantic depth implicit representation value of each reconstructed patient object raw data encoding feature vector, the information density or complexity carried by each patient data in the overall information is measured, and the importance and influence of each patient data in the overall information are quantified. Then, based on the sorting of the set of semantic depth implicit representation values ​​of patient object raw data, the set of patient object raw data encoding feature vectors is re-sorted to obtain the set of updated patient object raw data encoding feature vectors; that is, in a specific example of the present application, according to the semantic depth implicit representation value of each patient data, the feature order in the set is further adjusted, and the patient information structure is reorganized to ensure that key information is given priority in the subsequent contextual semantic association enhancement encoding process.

[0029] Among them, using a semantic reconstructor to perform feature reconstruction on each patient object raw data encoding feature vector in the set of patient object raw data encoding feature vectors to obtain a set of reconstructed patient object raw data encoding feature vectors, including: using a first weight parameter matrix and a second weight parameter matrix to perform continuous linear transformation on the patient object raw data encoding feature vectors to obtain the reconstructed patient object raw data encoding feature vectors.

[0030] More specifically, the semantic depth implicit representation value of each reconstructed patient object original data encoding feature vector in the set of reconstructed patient object original data encoding feature vectors is calculated to obtain a set of semantic depth implicit representation values ​​of patient object original data, including: calculating the square of the Euclidean norm of the reconstructed patient object original data encoding feature vector to obtain an initial semantic depth implicit representation value; calculating the ratio of the sum of the initial semantic depth implicit representation value and the difference between the initial semantic depth implicit representation value to obtain a semantic depth representation modulation coefficient; calculating the logarithmic function value of the semantic depth representation modulation coefficient with base 2 to obtain the semantic depth implicit representation value of the patient object original data.

[0031] Specifically, the bidirectional semantic association strengthening unit 332 is used to perform bidirectional contextual semantic association strengthening on the set of the updated patient object original data encoding feature vectors to obtain the set of the semantically strengthened patient object original data encoding feature vectors. In an embodiment of the present application, first, the set of the updated patient object original data encoding feature vectors is input into the first semantic encoder based on the forward LSTM model to obtain the set of forward patient object original data contextual semantic encoding feature vectors; and the set of the updated patient object original data encoding feature vectors is input into the second semantic encoder based on the backward LSTM model to obtain the set of backward patient object original data contextual semantic encoding feature vectors; that is, the bidirectional long short-term memory (LSTM) model is used to encode the set of updated patient object original data encoding feature vectors from the forward and backward directions respectively. Among them, the forward LSTM captures the information flow pattern from front to back, while the backward LSTM operates in reverse to identify the dependency from back to front. This dual-perspective encoding mechanism can capture the global dependency of patient information at different levels, and use it to strengthen the semantic feature expression of each patient data. Furthermore, the set of the forward patient object original data context semantic encoding feature vectors and the set of the backward patient object original data context semantic encoding feature vectors are fused to obtain the set of semantically enhanced patient object original data encoding feature vectors. Here, the bidirectional context semantic information of each patient data is integrated by fusing the encoding results of the bidirectional LSTM model to form a richer and more comprehensive set of semantically enhanced patient object original data encoding feature vectors, thereby achieving a deep understanding of patient information.

[0032] Among them, fusing the set of forward patient object original data context semantic coding feature vectors and the set of backward patient object original data context semantic coding feature vectors to obtain the set of semantically enhanced patient object original data coding feature vectors, includes: using a first weighting parameter and a second weighting parameter to weighted modulate the set of forward patient object original data context semantic coding feature vectors and the set of backward patient object original data context semantic coding feature vectors respectively to obtain a set of weighted modulated forward patient object original data context semantic coding feature vectors and a set of weighted modulated backward patient object original data context semantic coding feature vectors; calculating the corresponding points between the set of weighted modulated forward patient object original data context semantic coding feature vectors, the set of weighted modulated backward patient object original data context semantic coding feature vectors and the bias item to obtain the set of semantically enhanced patient object original data coding feature vectors.

[0033] In summary, in the above embodiment, the set of the patient object original data encoding feature vectors is semantically enhanced to obtain a set of semantically enhanced patient object original data encoding feature vectors, including: processing the set of the patient object original data encoding feature vectors with the following feature enhancement formula to obtain a set of semantically enhanced patient object original data encoding feature vectors, wherein the feature enhancement formula is:

[0034] T = {w1, w2, ..., w n}

[0035] h i =W1w i W2

[0036]

[0037] T f =M m v forward +H m v backward +b m

[0038] Where T represents the set of original data encoding feature vectors of the patient object, w1, w2, w i and w n represents the first, second, i-th and n-th patient object raw data encoding feature vectors in the set of the patient object raw data encoding feature vectors, respectively, n is the number of the patient object raw data encoding feature vectors, W1 and W2 represent the first weight parameter matrix and the second weight parameter matrix, respectively, h i is the feature vector encoded for the original data of the i-th reconstructed patient object, ||·|| represents the Euclidean norm of the calculated vector, log(·) represents the logarithmic function with base 2, e i represents the semantic depth implicit representation value of the original data of the i-th patient object, T order Represents the set of updated patient object raw data encoding feature vectors, v1, v2, v i and v n represent respectively the first, second, i-th and n-th updated patient object original data encoding feature vectors in the set of updated patient object original data encoding feature vectors, represents the forward LSTM model, represents the backward LSTM model, v forward represents the set of contextual semantic encoding feature vectors of the original data of the forward patient object, v backward represents the set of contextual semantic encoding feature vectors of the original data of the backward patient object, M m and H mdenote the first weighting parameter and the second weighting parameter respectively, b m is the bias term, T f A set of feature vectors is encoded for the semantically enhanced patient object raw data.

[0039] In particular, the multi-source data cross-modal fusion module 340 is used to perform cross-modal fusion on the set of the semantically enhanced patient object original data encoding feature vectors to obtain the patient object cross-modal multi-source data fusion feature vector. In a specific example of the present application, the set of the semantically enhanced patient object original data encoding feature vectors is input into a cross-modal deep semantic association feature mining module based on a multi-layer perceptron model to obtain the patient object cross-modal multi-source data fusion feature vector. It should be understood that in order to further integrate the patient's physical condition, test results, surgical information, and ASA grade and other information, the present application uses a multi-layer perceptron model to perform deep semantic association feature mining on the set of the semantically enhanced patient object original data encoding feature vectors, so as to utilize the nonlinear transformation capability of the multi-layer perceptron model to dig out the correlation dependency between the patient's multi-source data, extract and fuse the deep features of different modal data through layer-by-layer transmission and transformation, thereby realizing cross-modal feature fusion, generating a patient object cross-modal multi-source data fusion feature vector containing complementary information of multi-source data, so as to more comprehensively reflect the patient's overall health status and anesthesia risk.

[0040] In particular, the anesthesia risk probability estimation module 350 and the display module 360 ​​are used to determine the anesthesia risk probability based on the cross-modal multi-source data fusion feature vector of the patient object; and display the anesthesia risk probability on the display screen. In a specific example of the present application, the cross-modal multi-source data fusion feature vector of the patient object is input into the classifier-based anesthesia risk assessor to obtain the anesthesia risk probability. That is, by inputting the cross-modal multi-source data fusion feature vector of the patient object into the anesthesia risk assessor, the risk probability that the patient may face during anesthesia is predicted based on the cross-modal multi-source data fusion features of the patient object in the cross-modal multi-source data fusion feature vector of the patient object. Then, the anesthesia risk probability is displayed on the display screen. By displaying the anesthesia risk probability on the display screen, immediate and visual feedback can be provided to medical staff. In this way, more comprehensive and accurate decision support can be provided to anesthesiologists, significantly improving the efficiency and safety of perioperative patient anesthesia management.

[0041] As described above, a visual monitoring system 300 for anesthesia of a perioperative patient according to an embodiment of the present application can be implemented in various wireless terminals, such as a server having a visual monitoring algorithm for anesthesia of a perioperative patient. In one possible implementation, a visual monitoring system 300 for anesthesia of a perioperative patient according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the visual monitoring system 300 for anesthesia of a perioperative patient can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the visual monitoring system 300 for anesthesia of a perioperative patient can also be one of the many hardware modules of the wireless terminal.

[0042] Alternatively, in another example, the perioperative patient anesthesia visualization monitoring system 300 and the wireless terminal may also be separate devices, and the perioperative patient anesthesia visualization monitoring system 300 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0043] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A visual monitoring system for perioperative anesthesia, characterized in that: include: A patient object data collection module is used to collect the original data of the patient object to be anesthetized, wherein the original data includes basic physical indicators, laboratory indicators, operation information and ASA classification given by the doctor; A vectorization processing module, used for performing vectorization processing on each data item in the original data of the patient object to be anesthetized to obtain a set of encoding feature vectors of the original data of the patient object; A patient information context semantic enhancement module, used for performing semantic feature enhancement on the set of patient object original data encoding feature vectors to obtain a set of semantically enhanced patient object original data encoding feature vectors; A multi-source data cross-modal fusion module, used for cross-modal fusion of the set of semantically enhanced patient object original data encoding feature vectors to obtain a patient object cross-modal multi-source data fusion feature vector; an anesthesia risk probability estimation module, used to determine the anesthesia risk probability based on the cross-modal multi-source data fusion feature vector of the patient object; A display module, used for displaying the anesthesia risk probability on a display screen; Among them, the patient information context semantic enhancement module includes: a feature field effect distribution analysis unit, which is used to update the set of patient object original data encoding feature vectors based on the feature field effect distribution of the set of patient object original data encoding feature vectors to obtain a set of updated patient object original data encoding feature vectors; a bidirectional semantic association enhancement unit, which is used to perform bidirectional context semantic association enhancement on the set of updated patient object original data encoding feature vectors to obtain the set of semantically enhanced patient object original data encoding feature vectors.

2. A visual monitoring system for perioperative anesthesia according to claim 1, characterized in that: The characteristic field effect distribution analysis unit comprises: A semantic reconstruction subunit, configured to use a semantic reconstructor to perform feature reconstruction on each patient object original data encoding feature vector in the set of patient object original data encoding feature vectors to obtain a set of reconstructed patient object original data encoding feature vectors; A semantic depth representation subunit, used to calculate the semantic depth implicit representation value of each reconstructed patient object original data encoding feature vector in the set of reconstructed patient object original data encoding feature vectors to obtain a set of patient object original data semantic depth implicit representation values; The feature sorting subunit is used to re-sort the set of patient object original data encoding feature vectors based on the sorting of the set of patient object original data semantic depth implicit representation values ​​to obtain the set of updated patient object original data encoding feature vectors.

3. A visual monitoring system for perioperative anesthesia according to claim 2, characterized in that: The semantic reconstruction subunit is used to: The first weight parameter matrix and the second weight parameter matrix are used to continuously linearly transform the original data encoding feature vector of the patient object to obtain the reconstructed original data encoding feature vector of the patient object.

4. A visual monitoring system for perioperative anesthesia according to claim 3, characterized in that: The semantic depth representation subunit is used to: Calculating the square of the Euclidean norm of the reconstructed patient object original data encoding feature vector to obtain an initial semantic depth implicit representation value; Calculating a ratio of a sum of the initial semantic depth implicit representation value and a difference of the initial semantic depth implicit representation value to obtain a semantic depth representation modulation coefficient; The logarithmic function value with base 2 of the semantic depth representation modulation coefficient is calculated to obtain the semantic depth implicit representation value of the original data of the patient object.

5. A visual monitoring system for perioperative anesthesia according to claim 4, characterized in that: The bidirectional semantic association strengthening unit includes: A forward context semantic association encoding subunit, used for inputting the set of updated patient object original data encoding feature vectors into a first semantic encoder based on a forward LSTM model to obtain a set of forward patient object original data context semantic encoding feature vectors; A backward context semantic association encoding subunit, used for inputting the set of updated patient object original data encoding feature vectors into a second semantic encoder based on a backward LSTM model to obtain a set of backward patient object original data context semantic encoding feature vectors; A bidirectional fusion unit is used to fuse the set of forward patient object original data context semantic encoding feature vectors and the set of backward patient object original data context semantic encoding feature vectors to obtain the set of semantically enhanced patient object original data encoding feature vectors.

6. A visual monitoring system for perioperative anesthesia according to claim 5, characterized in that: The bidirectional fusion unit is used for: Using a first weighting parameter and a second weighting parameter, respectively, the set of forward patient object original data context semantic coding feature vectors and the set of backward patient object original data context semantic coding feature vectors are weighted modulated to obtain a set of weighted modulated forward patient object original data context semantic coding feature vectors and a set of weighted modulated backward patient object original data context semantic coding feature vectors; The set of contextual semantic coding feature vectors of the patient object original data before weighted modulation, the set of contextual semantic coding feature vectors of the patient object original data after weighted modulation, and the corresponding points between the bias items are calculated to obtain the set of semantically enhanced patient object original data coding feature vectors.

7. A visual monitoring system for perioperative anesthesia according to claim 6, characterized in that: The multi-source data cross-modal fusion module is used to: The set of feature vectors encoded with the original data of the semantically enhanced patient object is input into a cross-modal deep semantic association feature mining module based on a multi-layer perceptron model to obtain a cross-modal multi-source data fusion feature vector of the patient object.

8. A visual monitoring system for perioperative anesthesia according to claim 7, characterized in that: The anesthesia risk probability estimation module is used to: The cross-modal multi-source data fusion feature vector of the patient object is input into a classifier-based anesthesia risk assessor to obtain the anesthesia risk probability.

Images