Detection Of Apnea During Recovery From Anesthesia Based On Pulse Rate Variability

Objective:At first,this paper selected pulse rate variability indices from normal subjects which remains stable over time and can be distinguished between autonomy apnea and normal breathing,then in anesthesia recovery period patients,explore apnea influence on pulse rate variability,and pulse rate variability indices is used to detect the occurrence of apnea.Methods:We used polysomnography to collect pulse data of 45 healthy subjects in normal breathing state for 10 minutes,and discussed the characteristic values of pulse rate variability from the perspectives of time domain,frequency domain,nonlinear domain and time-frequency domain,and screened out 10 indicators of pulse variability independent of time duration.Then,the data of 45 healthy subjects under normal breathing and apnea were collected,and the differences of 10 characteristic values under normal breathing and apnea were calculated and compared,so as to screen out the indicators affected by apnea.Finally in anesthesia recovery room collected data of 52 patients with apnea,with the former part of indices of anesthesia recovery period in patients with apnea late prophase,during apnea,apnea and normal breathing condition were analyzed,and further to explore the anesthesia recovery apnea influence on patients of PRV,and reveals its impact on physiological.2.Based on the difference between apnea state and normal breathing state in patients recovering from anesthesia,the machine learning classification model of apnea was established by using pulse rate variability,and SMOTE,Borderline-SMOTE,and ADASYN three over-sampling methods were adopted to improve the generalization ability of the model.Results: In healthy subjects,10 PRV indices were screened out,which were not affected by duration,and could distinguish between normal breathing and apnea,namely MEAN、rMSSD、pNN50、nLF、nHF、LFHF、SD1、TF＿nLF、TF＿nHF、TF＿LFHF.The above 10 PRV indices were analyzed in the pre,during,and later stages of apnea and normal breathing state during anesthesia recovery period.Compared with normal breathing,only MEAN values were changed in the early stage of apnea,and there were differences in all 10 indices during apnea(P<0.05).Mean,RMSSD,PNN50,and SD1 values were different in the late stage of apnea(P<0.05).The sensitivity,specificity and accuracy of pulse rate variability in detecting apnea were 97.01%,85.54% and 89.31%Conclusion:Not only have we found a more sensitive indicator of PRV change between apnea and normal breathing during anesthesia convalescence,but also found that apnea during anesthesia convalescence can not only cause risk to patients,but also found that the effects of apnea on the body were not limited to the period of apnea,but also had a great impact in the short time after apnea.We can use pulse rate variability to detect the occurrence of apnea,and found that the over-sampling methods can improve the performance of the machine learning classification model.