Research And Implementation Of Sedation Depth Prediction Method Based On Machine Learning Sdn Based Flexible Deployment System Of Transition Technology

It is very important to monitor the depth of sedation during general anesthesia.On the one hand,too deep sedation will cause damage to the patient’s body.On the other hand,the depth of sedation is too shallow,which will bing the risk of perceived trauma to the patient during the operation.Bispectral index is a derivative index obtained from EEG signals.It is the most widely used index for sedation depth monitoring at present.However,according to the calculation of EEG,there is a 20-30s update delay in BIS,which affects the anesthesiologist’s judgment of the patient’s sedation state,increases the risk of the patient’s intraoperative awareness,and affects the patient’s postoperative recovery.Aiming at the problems of sedation monitoring in general anesthesia surgery,this thesis proposes a sedation depth prediction algorithm based on machine learning.The algorithm uses EEG sub parameters,historical BIS and intraoperative vital signs to predict the BIS value,so as to provide anesthesiologists with a more real-time reference of sedation state,so as to facilitate them to accurately adjust the anesthetic medication.Firstly,the noise of EEG signal is filtered by empirical mode decomposition algorithm,and the sub parameters are extracted.Then,the random forest algorithm was used to screen the sedation characteristic parameters,and the five characteristics with the highest correlation with the BIS index of sedation depth were obtained.Finally,the screened features and the corresponding BIS value are input into the long-term and short-term memory network for prediction.The experimental results show that the fitting accuracy between the predicted BIS value of the model and the real BIS value of patients is 0.93.Compared with multilayer perceptron and time-domain convolution network,the prediction accuracy of the algorithm is improved by 17.7%and 12.9%respectively.At the same time,the algorithm predicts that the BIS takes 0.32 seconds in 30 seconds,which is 0.2 seconds more than that of multi-layer perceptron and 2.12 seconds less than that of time-domain convolution network.Based on the above sedation depth prediction algorithm,this thesis designs and implements a sedation depth prediction system.The system first collects the data of medical equipment in the operating room,then analyzes the collected data,and finally displays the real-time data and the predicted sedation depth results.The data acquisition layer is responsible for collecting data and preprocessing the data to obtain the sedation characteristic screening data set;The data analysis layer needs to complete the screening of sedation parameters and the prediction of sedation depth;Finally,the result display layer subscribes the sedation depth prediction results through MQTT and displays them at the front end.In this thesis,the corresponding test cases are designed for each module of the system,and the results of the final test cases show that the system has achieved the established functions.