CN118197650B - An intelligent monitoring system for evaluating the safety of minimally invasive gynecological surgery - Google Patents

Abstract

An intelligent monitoring system for evaluating safety of gynecological minimally invasive surgery relates to the technical field of intelligent monitoring, and comprises a monitoring center, wherein the monitoring center is in communication connection with a feature extraction module, a data acquisition module, a real-time monitoring module, a multi-layer perception analysis module and an intelligent early warning module; the feature extraction module is used for extracting the associated features and the non-associated features of each operation flow sequence; the data acquisition module is used for acquiring monitoring data; the real-time monitoring module is used for generating a safety alarm signal of the gynecological minimally invasive surgery; the multi-layer perception analysis module acquires a predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence in the residual time period of the acquisition period based on the multi-layer perception early warning model; the intelligent early warning module is used for generating a surgical operation flow sequence review signal, so that the early sign of the complication before the condition of the complication is mature is predicted, and preventive measures are taken before the condition of the complication is mature.

Description

An intelligent monitoring system for evaluating the safety of minimally invasive gynecological surgery

Technical Field

The present invention relates to the field of intelligent monitoring technology, and in particular to an intelligent monitoring system for evaluating the safety of gynecological minimally invasive surgery.

Background technique

The prior art CN115565697A "A perioperative process monitoring and control system based on data analysis" solves the technical problem in the prior art that the preoperative, intraoperative and postoperative processes cannot be analyzed. It monitors the corresponding indicators of historical anesthesia operations, conducts risk assessment of anesthesia operations through indicator monitoring, and improves the monitoring pertinence of current anesthesia operations based on historical impact indicator analysis, thereby improving the safety and success of current operations; monitors the surgical indicators of the current anesthesia operation intraoperatively, determines the real-time impact of high-risk indicator parameters during the anesthesia operation, improves the accuracy of intraoperative monitoring, and improves the safety of anesthesia operations, thereby preventing abnormal anesthesia operations from reducing the success of the operation and affecting the subsequent treatment of patients;

The prior art CN117058854A "A fault monitoring and early warning system based on an integrated surgical power system" is used to solve the problem in the prior art that the integrated surgical power system cannot be monitored in real time before and during use, which leads to the integrated surgical power system being prone to failure, and then the operation cannot be carried out normally, causing difficulties or even errors in the surgical operation; the system can monitor the operating status of the running integrated surgical power system in real time, ensure the normal operating status of the integrated surgical power system, ensure the normal operation of the operation and improve the safety of the operation; the system can screen the integrated surgical power system multiple times before use, and obtain the integrated surgical power system with excellent comprehensive conditions for use in surgery, thereby reducing the probability of failure of the integrated surgical power system, further ensuring the normal operation of the operation and improving the safety of the operation;

In the existing intelligent monitoring technology for evaluating the safety of gynecological minimally invasive surgery, threshold monitoring technology is often used for various indicators during the operation. This is a static analysis method that fails to reflect the advantages of the online intelligent monitoring system. In addition, in the existing intelligent monitoring technology for evaluating the safety of gynecological minimally invasive surgery, indicators that are closely related to complications during the patient's operation are not differentiated and analyzed, resulting in low accuracy of intelligent monitoring of the safety of gynecological minimally invasive surgery. There are weak or difficult-to-identify signs during the operation. If the operation process is not reviewed or further detailed during the operation, this may lead to the occurrence of intraoperative complications. Based on the existing threshold monitoring technology, early signs before complications occur during the operation are often difficult to detect, because the signs of complications during the operation may be weak or difficult to identify, which may lead to the inability to take preventive measures before the conditions for complications are ripe. How to solve the above technical difficulties is a problem that we urgently need to solve.

Summary of the invention

In order to solve the above technical problems, the object of the present invention is to provide an intelligent monitoring system for evaluating the safety of gynecological minimally invasive surgery, including a monitoring center, wherein the monitoring center is communicatively connected with a feature extraction module, a data acquisition module, a real-time monitoring module, a multi-layer perception analysis module and an intelligent early warning module;

The feature extraction module is used to obtain the standardized operation process of gynecological minimally invasive surgery, and extract the correlation features and non-correlation features of each operation process sequence in the standardized operation process;

The data acquisition module is used to collect monitoring data of each operation process sequence and mark the collection time and set the collection cycle;

The real-time monitoring module is used to determine whether each operation process sequence is within the corresponding qualified threshold interval, and generate a gynecological minimally invasive surgery safety alarm signal according to the determination result;

The multi-layer perception analysis module constructs a multi-layer perception warning model based on deep learning, and obtains the predicted data time series sequence corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle based on the multi-layer perception warning model;

The intelligent early warning module analyzes the predicted data time series corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle, and generates a surgical operation process sequence review signal according to the analysis result.

Furthermore, the feature extraction module obtains the standardized operation process of gynecological minimally invasive surgery, and the process of extracting the associated features and non-associated features of each operation process sequence in the standardized operation process includes:

Extract the operation process sequence according to the standardized operation process of gynecological minimally invasive surgery, and extract the associated features of the operation process sequence;

Big data technology is used to retrieve all monitoring indicators and several historical medical records corresponding to gynecological minimally invasive surgery, and all monitoring indicators corresponding to gynecological minimally invasive surgery are marked as judgment and research indicators. Based on several historical medical records, the numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have no records of complications during the operation are collected. At the same time, the numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have records of complications during the operation are collected. The numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have no records of complications during the operation are compared with the numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have records of complications during the operation;

If the numerical range of the judgment and research indicators of each operation process sequence corresponding to the patients who have completed gynecological minimally invasive surgery in the past and have no complication records during the surgery is consistent with the numerical range of the judgment and research indicators of each operation process sequence corresponding to the patients who have completed gynecological minimally invasive surgery in the past and have complications records during the surgery, then obtain the monitoring indicators corresponding to the judgment and research indicators with the same numerical range of each operation process sequence and mark them as unrelated monitoring indicators;

If the numerical range of the judgment and research indicators of each operation process sequence corresponding to patients who have completed gynecological minimally invasive surgery with no record of complications during the operation is inconsistent with the numerical range of the judgment and research indicators of each operation process sequence corresponding to patients who have completed gynecological minimally invasive surgery with a record of complications during the operation, then the monitoring indicators corresponding to the judgment and research indicators with inconsistent numerical ranges of each operation process sequence are obtained and marked as associated monitoring indicators.

Furthermore, the data collection module collects monitoring data of each operation process sequence and marks the collection time. The process of setting the collection cycle includes:

Setting data monitoring points, wherein the data monitoring points collect monitoring data corresponding to the unrelated monitoring indicators of each operation process sequence according to the unrelated monitoring indicators of each operation process sequence, mark the collection time, and set the collection cycle;

The data monitoring points collect monitoring data corresponding to the associated monitoring indicators of each operation process sequence according to the associated monitoring indicators of each operation process sequence, mark the collection time, and set the collection cycle.

Furthermore, the process of the real-time monitoring module determining whether each operation process sequence is within the corresponding qualified threshold interval and generating a gynecological minimally invasive surgery safety alarm signal according to the determination result includes:

Preset the qualified threshold intervals corresponding to the unrelated monitoring indicators of each operation process sequence of gynecological minimally invasive surgery, and the qualified threshold intervals corresponding to the related monitoring indicators of each operation process sequence;

Acquire monitoring data corresponding to unrelated monitoring indicators of several operation process sequences and monitoring data corresponding to related monitoring indicators of several operation process sequences collected at data monitoring points within a collection period, and determine whether the monitoring data corresponding to the unrelated monitoring indicators and the monitoring data corresponding to the related monitoring indicators are within corresponding qualified threshold intervals;

If the monitoring data is in the corresponding qualified threshold interval, the monitoring data corresponding to the associated monitoring indicators of several operation process sequences are subjected to trend characterization analysis;

If there is monitoring data that is not within the corresponding qualified threshold interval, a gynecological minimally invasive surgery safety alarm signal is generated and sent to the monitoring center.

Furthermore, the process of constructing a multi-layer perception warning model based on deep learning by the multi-layer perception analysis module includes:

Building a multi-layer perception warning model based on deep learning, obtaining training data through a number of historical medical records, and using the training data to train the multi-layer perception warning model;

The training data is input into the multi-layer perception warning model for training until the loss function training is stable, and the model parameters are saved. The multi-layer perception warning model is tested through a test set until it meets the preset requirements, and the multi-layer perception warning model is output.

Furthermore, the process of the multi-layer perception analysis module obtaining the predicted data time series corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle includes:

Acquire monitoring data corresponding to associated monitoring indicators of several operation process sequences collected within a collection period, extract temporal features and spatial features of the monitoring data corresponding to associated monitoring indicators of several operation process sequences, and generate temporal and spatial feature sequences of the monitoring data of several operation process sequences;

The monitoring data of several operation process sequences and the spatiotemporal feature sequences of the monitoring data of several operation process sequences are input into the multi-layer perception early warning model, and the multi-layer perception early warning model is used to generate a prediction data time series sequence corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle.

Furthermore, the multi-layer perception analysis module extracts temporal features and spatial features of the monitoring data corresponding to the associated monitoring indicators of the plurality of operation process sequences respectively, including:

Obtain the implementation order and implementation relationship between each operation process sequence, take each operation process sequence as a node of a topological directed graph, take the implementation order and implementation relationship between each operation process sequence as a connection relationship between nodes, and construct a topological directed graph;

Acquire monitoring data corresponding to associated monitoring indicators of several operation process sequences collected during the collection period, splice the monitoring data corresponding to the associated monitoring indicators collected at each moment during the monitoring time period of several operation process sequences according to the time series relationship, generate a two-dimensional feature matrix, construct a temporal convolutional neural network to learn the two-dimensional feature matrices of several operation process sequences, and obtain monitoring data change characteristics according to the temporal convolutional neural network after learning;

A graph attention network is constructed to learn the topological directed graph. The change characteristics of the monitoring data of several operation process sequences are input into the graph attention network. The attention weights of the adjacent nodes of each node in the topological directed graph are obtained through the attention mechanism. The neighbor aggregation mechanism of the graph attention network is used to generate the spatiotemporal characteristics of the monitoring data of each node according to the attention weight and the temporal characteristics of the monitoring data of each node.

Furthermore, the intelligent early warning module analyzes the predicted data time series corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle, and generates a surgical operation process sequence review signal according to the analysis result, including:

Obtain the qualified threshold interval corresponding to the associated monitoring indicator of each operation process sequence, obtain the numerical fluctuation coefficient of the predicted data time series sequence corresponding to the associated monitoring indicator of each operation process sequence, and the frequency that the numerical value of the predicted data time series sequence corresponding to the associated monitoring indicator of each operation process sequence is not within the corresponding qualified threshold interval, and obtain the early warning characterization coefficient of the predicted data time series sequence according to the numerical fluctuation coefficient and the frequency;

A threshold value of the early warning characterization coefficient is preset. When the early warning characterization coefficient of the predicted data time series sequence corresponding to the associated monitoring indicator of the operation process sequence is less than the threshold value of the early warning characterization coefficient, a normal operation signal is generated and sent to the monitoring center;

When the early warning characterization coefficient of the predicted data time series sequence corresponding to the associated monitoring indicator of the operation process sequence is greater than or equal to the early warning characterization coefficient threshold, a surgical operation process sequence review signal is generated and sent to the monitoring center.

Compared with the prior art, the beneficial effects of the present invention are as follows: the associated features and non-associated features of each operation process sequence in the standardized operation process are extracted through the feature extraction module, and firstly, whether each operation process sequence is located in the corresponding qualified threshold interval is judged through the real-time monitoring module, and a gynecological minimally invasive surgery safety alarm signal is generated according to the judgment result, and online threshold monitoring is performed, and then, based on the online threshold monitoring, a multi-layer perception early warning model is established through the multi-layer perception analysis module, and the predicted data time series sequence corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle is obtained based on the multi-layer perception early warning model, and the predicted data time series sequence corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle is analyzed, and a surgical operation process sequence review signal is generated according to the analysis result, so as to realize the prediction of early signs before the conditions for the occurrence of complications are ripe, so as to take preventive measures before the conditions for the occurrence of complications are ripe.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of an intelligent monitoring system for evaluating the safety of minimally invasive gynecological surgery according to an embodiment of the present application.

Detailed ways

The following is a clear and complete description of the technical solutions in the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

As shown in FIG1 , an intelligent monitoring system for evaluating the safety of gynecological minimally invasive surgery includes a monitoring center, wherein the monitoring center is communicatively connected to a feature extraction module, a data acquisition module, a real-time monitoring module, a multi-layer perception analysis module, and an intelligent early warning module;

The feature extraction module is used to obtain the standardized operation process of gynecological minimally invasive surgery, and extract the correlation features and non-correlation features of each operation process sequence in the standardized operation process;

The data acquisition module is used to collect monitoring data of each operation process sequence and mark the collection time and set the collection cycle;

The real-time monitoring module is used to determine whether each operation process sequence is within the corresponding qualified threshold interval, and generate a gynecological minimally invasive surgery safety alarm signal according to the determination result;

The multi-layer perception analysis module constructs a multi-layer perception warning model based on deep learning, and obtains the predicted data time series sequence corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle based on the multi-layer perception warning model;

The intelligent early warning module analyzes the predicted data time series corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle, and generates a surgical operation process sequence review signal according to the analysis result.

It should be further explained that, in the specific implementation process, the feature extraction module obtains the standardized operation process of gynecological minimally invasive surgery, and the process of extracting the associated features and non-associated features of each operation process sequence in the standardized operation process includes:

Extract the operation process sequence according to the standardized operation process of gynecological minimally invasive surgery, and extract the associated features of the operation process sequence;

Big data technology is used to retrieve all monitoring indicators and several historical medical records corresponding to gynecological minimally invasive surgery, and all monitoring indicators corresponding to gynecological minimally invasive surgery are marked as judgment and research indicators. Based on several historical medical records, the numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have no records of complications during the operation are collected. At the same time, the numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have records of complications during the operation are collected. The numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have no records of complications during the operation are compared with the numerical ranges of the judgment and research indicators of each operation process sequence corresponding to patients who have historically completed gynecological minimally invasive surgery and have records of complications during the operation;

If the numerical range of the judgment and research indicators of each operation process sequence corresponding to the patients who have completed gynecological minimally invasive surgery in the past and have no complication records during the surgery is consistent with the numerical range of the judgment and research indicators of each operation process sequence corresponding to the patients who have completed gynecological minimally invasive surgery in the past and have complications records during the surgery, then obtain the monitoring indicators corresponding to the judgment and research indicators with the same numerical range of each operation process sequence and mark them as unrelated monitoring indicators;

If the numerical range of the judgment and research indicators of each operation process sequence corresponding to patients who have completed gynecological minimally invasive surgery with no record of complications during the operation is inconsistent with the numerical range of the judgment and research indicators of each operation process sequence corresponding to patients who have completed gynecological minimally invasive surgery with a record of complications during the operation, then the monitoring indicators corresponding to the judgment and research indicators with inconsistent numerical ranges of each operation process sequence are obtained and marked as associated monitoring indicators.

The operating sequence of minimally invasive gynecological surgery includes:

Preparation steps: The doctor and surgical team will prepare the necessary instruments, equipment and drugs, and the patient will be taken to the operating room and properly positioned on the operating table;

Anesthesia and analgesia steps: The anesthesiologist will anesthetize the patient to ensure that the patient does not feel pain during the operation. Analgesia measures will also be taken before the operation to reduce pain and discomfort after the operation;

Surgical site preparation steps: The doctor will disinfect and drape the surgical site to ensure the cleanliness and sterility of the surgical field. In special cases, special surgical site preparation may be required, such as vaginal douching;

Small incision or endoscopic access steps: For laparoscopic surgery, the doctor will make a small incision in the patient's abdomen and insert a laparoscope to observe and operate the internal abdominal organs. For hysteroscopic surgery, the doctor will insert a hysteroscope into the uterus through the vagina to perform endoscopic examination and operation procedures;

Surgical operation process: The doctor uses laparoscope or hysteroscope to perform relevant tissue removal steps, repair steps or reconstruction steps according to the patient's condition and surgical needs. During the operation, the doctor may use various minimally invasive surgical instruments, such as forceps, scissors, suction devices, etc.;

Hemostasis and suturing steps: During the operation, the doctor will perform necessary hemostasis treatment to reduce the risk of postoperative bleeding. After the operation, the doctor will suture or fit the surgical site to ensure that the wound heals well;

Postoperative treatment steps: After the operation, the patient will be sent to the recovery room for observation and monitoring until the anesthesia effect completely subsides and the vital signs are stable; the doctor will perform appropriate bandaging and care for the surgical site;

The operation process of each step in the operation sequence will directly affect the difficulty and quality of completing the next step. If there are faint or difficult-to-identify signs during the operation, if the operation process is not reviewed or further detailed during the operation, this may lead to the occurrence of intraoperative complications.

Monitoring indicators include but are not limited to:

Vital signs data: including basic physiological indicators such as the patient's heart rate, respiratory rate, and body temperature. These indicators reflect the patient's overall physiological state and are of great significance for evaluating the safety of surgery;

Blood pressure data: including systolic pressure, diastolic pressure and mean arterial pressure and other blood pressure parameters. Monitoring blood pressure can timely detect abnormal conditions in the circulatory system, such as hypotension, hypertension, etc.;

Respiratory parameter data: including the patient's respiratory rate, tidal volume, breathing depth and other respiratory-related indicators. These data can be used to evaluate the patient's respiratory function and ventilation status;

Anesthesia monitoring data: including the use of anesthetic drugs, anesthesia depth monitoring indicators, etc. These data are crucial for evaluating the anesthesia effect and safety of patients during surgery;

Blood oxygen saturation data: By monitoring the patient's blood oxygen saturation, problems such as poor oxygenation or hypoxemia can be discovered in a timely manner to ensure that the patient's respiratory function is normal;

Surgical operation data: including the use of surgical instruments, operation steps and time during the operation, etc. These data can be used to evaluate the standardization and safety of surgical operations.

Complication monitoring data: Monitoring possible complications during surgery, such as bleeding, instrument damage, organ damage, etc. Timely detection and treatment of these complications are crucial to ensuring surgical safety.

It should be further explained that, in the specific implementation process, the data acquisition module collects the monitoring data of each operation process sequence and marks the collection time. The process of setting the collection cycle includes:

Setting data monitoring points, wherein the data monitoring points collect monitoring data corresponding to the unrelated monitoring indicators of each operation process sequence according to the unrelated monitoring indicators of each operation process sequence, mark the collection time, and set the collection cycle;

The data monitoring points collect monitoring data corresponding to the associated monitoring indicators of each operation process sequence according to the associated monitoring indicators of each operation process sequence, mark the collection time, and set the collection cycle.

It should be further explained that, in the specific implementation process, the real-time monitoring module determines whether each operation process sequence is within the corresponding qualified threshold interval, and generates a gynecological minimally invasive surgery safety alarm signal according to the judgment result, including:

Preset the qualified threshold intervals corresponding to the unrelated monitoring indicators of each operation process sequence of gynecological minimally invasive surgery, and the qualified threshold intervals corresponding to the related monitoring indicators of each operation process sequence;

Acquire monitoring data corresponding to unrelated monitoring indicators of several operation process sequences and monitoring data corresponding to related monitoring indicators of several operation process sequences collected at data monitoring points within a collection period, and determine whether the monitoring data corresponding to the unrelated monitoring indicators and the monitoring data corresponding to the related monitoring indicators are within corresponding qualified threshold intervals;

If the monitoring data is in the corresponding qualified threshold interval, the monitoring data corresponding to the associated monitoring indicators of several operation process sequences are subjected to trend characterization analysis;

If there is monitoring data that is not within the corresponding qualified threshold interval, a gynecological minimally invasive surgery safety alarm signal is generated and sent to the monitoring center.

It should be further explained that, in the specific implementation process, the process of the multi-layer perception analysis module constructing a multi-layer perception warning model based on deep learning includes:

Building a multi-layer perception warning model based on deep learning, obtaining training data through a number of historical medical records, and using the training data to train the multi-layer perception warning model;

The training data is input into the multi-layer perception warning model for training until the loss function training is stable, and the model parameters are saved. The multi-layer perception warning model is tested through a test set until it meets the preset requirements, and the multi-layer perception warning model is output.

It should be further explained that, in the specific implementation process, the process of the multi-layer perception analysis module obtaining the predicted data time series sequence corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle includes:

Acquire monitoring data corresponding to associated monitoring indicators of several operation process sequences collected within a collection period, extract temporal features and spatial features of the monitoring data corresponding to associated monitoring indicators of several operation process sequences, and generate temporal and spatial feature sequences of the monitoring data of several operation process sequences;

The monitoring data of several operation process sequences and the spatiotemporal feature sequences of the monitoring data of several operation process sequences are input into the multi-layer perception early warning model, and the multi-layer perception early warning model is used to generate a prediction data time series sequence corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle.

It should be further explained that, in the specific implementation process, the multi-layer perception analysis module extracts the temporal features and spatial features of the monitoring data corresponding to the associated monitoring indicators of several operation process sequences respectively, including:

Obtain the implementation order and implementation relationship between each operation process sequence, take each operation process sequence as a node of a topological directed graph, take the implementation order and implementation relationship between each operation process sequence as the connection relationship between the nodes, and construct a topological directed graph;

Acquire monitoring data corresponding to associated monitoring indicators of several operation process sequences collected during the collection period, splice the monitoring data corresponding to the associated monitoring indicators collected at each moment during the monitoring time period of several operation process sequences according to the time series relationship, generate a two-dimensional feature matrix, construct a temporal convolutional neural network to learn the two-dimensional feature matrices of several operation process sequences, and obtain monitoring data change characteristics according to the temporal convolutional neural network after learning;

A graph attention network is constructed to learn the topological directed graph. The change characteristics of the monitoring data of several operation process sequences are input into the graph attention network. The attention weights of the adjacent nodes of each node in the topological directed graph are obtained through the attention mechanism. The neighbor aggregation mechanism of the graph attention network is used to generate the spatiotemporal characteristics of the monitoring data of each node according to the attention weight and the temporal characteristics of the monitoring data of each node.

It should be further explained that, in the specific implementation process, the monitoring data corresponding to the associated monitoring indicators collected at each moment during the monitoring time period of several operation process sequences are spliced according to the time series relationship, and the formula for generating a two-dimensional feature matrix is:

=concact(Q1j,Q2j,...Qij,...,QNj), i=1,2,...,N;

j＝1,2,...,K;

Among them, Qij represents the characteristic data matrix corresponding to the monitoring data corresponding to the associated monitoring indicator of the j-th operation process sequence at the i-th moment, N represents the total number of monitoring moments of the j-th operation process sequence, R represents the number of associated monitoring indicator types, K represents the total number of operation process sequences, and concact represents the splicing of the characteristic data matrices at each moment according to the time series relationship. is the two-dimensional data matrix of the constructed j-th operation process sequence;

A temporal convolutional neural network is constructed to learn the two-dimensional feature matrix of several operation process sequences. The formula for obtaining the characteristics of monitoring data changes based on the temporal convolutional neural network after learning is as follows:

=TCN（ );

in, represents the monitoring data change characteristics of the j-th operation process sequence, and TCN() represents the function of the temporal convolutional neural network.

It should be further explained that in the specific implementation process, the attention weight of the neighboring nodes of each node in the topological directed graph is obtained through the attention mechanism, and the neighbor aggregation mechanism of the graph attention network is used to generate the spatiotemporal feature sequence of the monitoring data of each node according to the attention weight and the time characteristics of the monitoring data of each node. The specific process is:

;

;

;

...

;

in, express The corresponding adjacency matrix, ELU and Tanh represent activation functions, and FULLY represents the fully connected layer of the graph attention network. and Respectively represent the weight matrix and bias corresponding to the x+1th iteration, Represents the spatiotemporal feature sequence of the j-th operation process sequence after the x+1th iteration.

It should be further explained that, in the specific implementation process, the intelligent early warning module analyzes the predicted data time series corresponding to the associated monitoring indicators of each operation process sequence in the remaining time period of the acquisition cycle, and generates a surgical operation process sequence review signal according to the analysis results, including:

Obtain the qualified threshold interval corresponding to the associated monitoring indicator of each operation process sequence, obtain the numerical fluctuation coefficient of the predicted data time series sequence corresponding to the associated monitoring indicator of each operation process sequence, and the frequency that the numerical value of the predicted data time series sequence corresponding to the associated monitoring indicator of each operation process sequence is not within the corresponding qualified threshold interval, and obtain the early warning characterization coefficient of the predicted data time series sequence according to the numerical fluctuation coefficient and the frequency;

A threshold value of the early warning characterization coefficient is preset. When the early warning characterization coefficient of the predicted data time series sequence corresponding to the associated monitoring indicator of the operation process sequence is less than the threshold value of the early warning characterization coefficient, a normal operation signal is generated and sent to the monitoring center;

When the early warning characterization coefficient of the predicted data time series sequence corresponding to the associated monitoring indicator of the operation process sequence is greater than or equal to the early warning characterization coefficient threshold, a surgical operation process sequence review signal is generated and sent to the monitoring center.

It should be further explained that, during the specific implementation process, the surgical operation process sequence review signal indicates that the various monitoring data of the current operation process sequence are qualified, but the doctor performing the operation or other experienced doctors on site need to review the patient's current completed operation process sequence. Through repeated inspections, doctors can promptly discover potential problems or errors and correct them in a timely manner to ensure the safety of the surgical process. This is crucial to avoiding unexpected events or complications during surgery, helping to ensure that patients receive high-quality medical treatment, minimize surgical risks, and protect patients' lives and health.

It should be further explained that, in the specific implementation process, the calculation formula for the numerical fluctuation coefficient of the predicted data time series corresponding to the associated monitoring indicators of each operation process sequence is:

;

in, The numerical fluctuation coefficient of the predicted data time series corresponding to the associated monitoring indicator of the jth operation process sequence; represents the value corresponding to the predicted data at the tth moment corresponding to the associated monitoring indicator of the jth operation process sequence; represents the average value of the predicted data corresponding to the associated monitoring indicator of the j-th operation process sequence; N represents the total number of monitoring moments of the j-th operation process sequence;

The calculation formula for obtaining the early warning characterization coefficient of the forecast data time series according to the numerical fluctuation coefficient and the frequency is:

;

in, represents the early warning characterization coefficient of the prediction data corresponding to the associated monitoring indicator of the j-th operation process sequence, The frequency that the value of the predicted data time series corresponding to the associated monitoring indicator of the j-th operation process sequence is not within the corresponding qualified threshold interval, Represents the conversion coefficient.

The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims (1)

1. The intelligent monitoring system for evaluating the safety of the gynecological minimally invasive surgery comprises a monitoring center, and is characterized in that the monitoring center is in communication connection with a feature extraction module, a data acquisition module, a real-time monitoring module, a multi-layer perception analysis module and an intelligent early warning module;

The feature extraction module is used for obtaining a standardized operation flow of the gynecological minimally invasive surgery and extracting associated features and non-associated features of each operation flow sequence in the standardized operation flow;

the feature extraction module acquires a standardized operation flow of the gynecological minimally invasive surgery, and the process for extracting the associated features and the non-associated features of each operation flow sequence in the standardized operation flow comprises the following steps:

extracting an operation flow sequence according to a standardized operation flow of the gynecological minimally invasive surgery, and extracting associated features of the operation flow sequence;

Searching all monitoring indexes corresponding to the gynecological minimally invasive surgery and a plurality of historical medical records by utilizing a big data technology, marking all monitoring indexes corresponding to the gynecological minimally invasive surgery as judging and grinding indexes, collecting the numerical ranges of judging and grinding indexes of each operation flow sequence corresponding to a patient who has historically completed the gynecological minimally invasive surgery and has no complication record in operation according to the plurality of historical medical records, collecting the numerical ranges of judging and grinding indexes of each operation flow sequence corresponding to a patient who has historically completed the gynecological minimally invasive surgery and has produced the complication record in operation, and comparing the numerical ranges of judging and grinding indexes of each operation flow sequence corresponding to a patient who has historically completed the gynecological minimally invasive surgery and has no complication record in operation with the numerical ranges of judging and grinding indexes of each operation flow sequence corresponding to a patient who has historically completed the gynecological minimally invasive surgery and has produced the complication record in operation;

If the numerical range of the judging and researching index of each operation flow sequence corresponding to the patient who completes the gynecological minimally invasive surgery and has no complication record in the operation is consistent with the numerical range of the judging and researching index of each operation flow sequence corresponding to the patient who completes the gynecological minimally invasive surgery and has the complication record in the operation, acquiring a monitoring index corresponding to the judging and researching index with consistent numerical range of each operation flow sequence, and marking the monitoring index as an unassociated monitoring index;

If the numerical range of the judging and researching index of each operation flow sequence corresponding to the patient who completes the gynecological minimally invasive surgery and has no complication record in the operation is inconsistent with the numerical range of the judging and researching index of each operation flow sequence corresponding to the patient who completes the gynecological minimally invasive surgery and has the complication record in the operation, acquiring a monitoring index corresponding to the judging and researching index inconsistent with the numerical range of each operation flow sequence, and marking the monitoring index as an associated monitoring index;

The data acquisition module is used for acquiring monitoring data of each operation flow sequence, marking acquisition time and setting an acquisition period;

The data acquisition module acquires monitoring data of each operation flow sequence and marks acquisition time, and the process of setting the acquisition period comprises the following steps:

Setting a data monitoring point, wherein the data monitoring point acquires the monitoring data mark acquisition time corresponding to the unassociated monitoring index of each operation flow sequence according to the unassociated monitoring index of each operation flow sequence, and sets an acquisition period;

the data monitoring point location acquires the monitoring data mark acquisition time corresponding to the associated monitoring index of each operation flow sequence according to the associated monitoring index of each operation flow sequence, and sets an acquisition period;

the real-time monitoring module is used for judging whether each operation flow sequence is located in a corresponding qualified threshold interval or not, and generating a gynecological minimally invasive surgery safety alarm signal according to a judgment result;

the process of judging whether each operation flow sequence is located in the corresponding qualified threshold interval by the real-time monitoring module and generating the gynecological minimally invasive surgery safety alarm signal according to the judgment result comprises the following steps:

Presetting a corresponding qualification threshold interval of unassociated monitoring indexes of each operation flow sequence of the gynecological minimally invasive surgery and a corresponding qualification threshold interval of associated monitoring indexes of each operation flow sequence;

Acquiring monitoring data corresponding to unassociated monitoring indexes of a plurality of operation flow sequences and monitoring data corresponding to associated monitoring indexes of a plurality of operation flow sequences, which are acquired by data monitoring points in an acquisition period, and judging whether the monitoring data corresponding to unassociated monitoring indexes and the monitoring data corresponding to associated monitoring indexes are located in corresponding qualified threshold intervals or not;

If the monitoring data are located in the corresponding qualified threshold interval, trend characterization analysis is carried out on the monitoring data corresponding to the associated monitoring indexes of the operation flow sequences;

if monitoring data which is not located in the corresponding qualified threshold interval exists, generating a gynecological minimally invasive surgery safety alarm signal and sending the safety alarm signal to a monitoring center;

The multi-layer perception analysis module builds a multi-layer perception early warning model based on deep learning, and obtains a predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence in the residual time period of the acquisition period based on the multi-layer perception early warning model;

The process for constructing the multi-layer perception early warning model based on the deep learning by the multi-layer perception analysis module comprises the following steps:

Constructing a multi-layer perception early warning model based on deep learning, acquiring training data through a plurality of historical medical records, and training the multi-layer perception early warning model by utilizing the training data;

Inputting training data into the multi-layer perception early-warning model for training until the loss function training is stable, storing model parameters, testing the multi-layer perception early-warning model through a test set until the multi-layer perception early-warning model meets preset requirements, and outputting the multi-layer perception early-warning model;

The process of obtaining the predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence in the residual time period of the acquisition period by the multi-layer perception analysis module comprises the following steps:

Acquiring the monitoring data corresponding to the relevant monitoring indexes of the collected operation flow sequences in the collection period, respectively extracting the time characteristics and the space characteristics of the monitoring data corresponding to the relevant monitoring indexes of the operation flow sequences, and generating a time-space characteristic sequence of the monitoring data of the operation flow sequences;

inputting the monitoring data of a plurality of operation flow sequences and the time-space characteristic sequences of the monitoring data of the operation flow sequences into a multi-layer perception early warning model, and generating a predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence in the residual time period of the acquisition period through the multi-layer perception early warning model;

The process of extracting the time features and the space features of the monitoring data corresponding to the associated monitoring indexes of the operation flow sequences by the multi-layer perception analysis module comprises the following steps:

Acquiring an implementation sequence and an implementation relation among the operation flow sequences, taking the operation flow sequences as nodes of the topology directed graph, and taking the implementation sequence and the implementation relation among the operation flow sequences as a connection relation among the nodes to construct the topology directed graph;

Acquiring monitoring data corresponding to the associated monitoring indexes of a plurality of operation flow sequences acquired in an acquisition period, splicing the monitoring data corresponding to the associated monitoring indexes at each moment acquired in a monitoring time period of the plurality of operation flow sequences according to a time sequence relation, generating a two-dimensional feature matrix, constructing a time convolution neural network to learn the two-dimensional feature matrix of the plurality of operation flow sequences, and acquiring monitoring data change features according to the time convolution neural network after learning;

Constructing a graph attention network to learn a topological directed graph, inputting the change characteristics of the monitoring data of a plurality of operation flow sequences into the graph attention network, acquiring the attention weight of the adjacent node of each node in the topological directed graph to the self through an attention mechanism, and generating the time-space characteristics of the monitoring data of each node according to the attention weight and the time characteristics of the monitoring data of each node by utilizing a neighbor aggregation mechanism of the graph attention network;

the intelligent early warning module analyzes the predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence in the residual time period of the acquisition period, and generates an operation flow sequence review signal according to the analysis result;

The intelligent early warning module analyzes the predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence in the residual time period of the acquisition period, and the process of generating the operation flow sequence review signal according to the analysis result comprises the following steps:

Acquiring a qualified threshold interval corresponding to an associated monitoring index of each operation flow sequence, acquiring a numerical fluctuation coefficient of a predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence and a frequency of which the numerical value of the predicted data time sequence corresponding to the associated monitoring index of each operation flow sequence is not located in the corresponding qualified threshold interval, and acquiring an early warning characterization coefficient of the predicted data time sequence according to the numerical fluctuation coefficient and the frequency;

Presetting an early warning characterization coefficient threshold, and generating a surgery normal signal and sending the surgery normal signal to a monitoring center when the early warning characterization coefficient of a predicted data time sequence corresponding to an associated monitoring index of an operation flow sequence is smaller than the early warning characterization coefficient threshold;

When the early warning characterization coefficient of the predicted data time sequence corresponding to the associated monitoring index of the operation flow sequence is larger than or equal to the threshold value of the early warning characterization coefficient, generating a surgical operation flow sequence review signal and sending the surgical operation flow sequence review signal to a monitoring center.