Feature Optimization And Respiratory Rate Monitoring Based On Tracheal Sounds

Objective: Respiratory flow rate estimation using respiratory sounds received considerable attention.Several tracheal sound features have been investigated to establish a model that best describes flow-sound relationship.There are still some shortcomings in the method of monitoring respiratory rate in clinic.In this study,we aimed to validate the relationship between the flow rate and the features of tracheal sounds and select an optimal feature set to estimate flow rate with minimum estimation error.Then based on the optimal feature,respiratory rate was monitored under different respiratory conditions.Methods: 1.Tracheal sounds and flow rate were acquired from nine healthy volunteers under resting states.Seven types of tracheal sound features were extracted for flow rate estimation.Effects of respiratory flow rate estimation using these features were evaluated using Bland-Altman analysis,estimation error and tracheal sounds feature calculation time.Besides,expiratory and inspiratory phases divided estimation error was compared with united estimation error.2.Tracheal sounds and nasal flow pressure signal were acquired from 46 healthy volunteers.Five methods were used to calculate the respiratory rate based on tracheal sounds feature under rate stable states.The method of estimating the number of breaths was used to calculate the respiratory rate under rate changes states.Respiratory rate based on nasal flow pressure signal was taken as reference respiratory rate.Effects of monitoring respiratory rate based on tracheal sound feature were evaluated using linear regression analysis,accuracy and precision,success rate and Bland-Altman analysis.Results: 1.For each set of respiratory sound features,Bland-Altman analysis showed the satisfactory correlation and agreement between the actual flow rate and the estimated flow rate.If the estimation effect was considered only from estimation error,Log Variance?100ms,50%?was the optimal feature for tracheal sound estimating respiratory flow rate,if the estimation effect was considered form estimation error and the calculation time of features two aspects,Log Envelope was the optimal feature for tracheal sound estimating respiratory flow rate,at the same time,united estimation error as lower than divided estimation error.2.Combining linear regression analysis,accuracy and precision,success rate and Bland-Altman analysis,results of respiratory rate based on tracheal sound feature was as follows: the results based on AMDF₁method were better than those of other methods under rate stable states.Under rate obviously changes states,there was a obvious difference between the results of respiratory rate obtained by AMDF₁ method and reference respiratory rate for four different respiratory conditions,and it cannot reflect the change of respiratory rate.Therefore,the method of estimating the number of breaths was used to calculate the respiratory rate based on tracheal sound features in the case of respiratory rate obvious changes,the respiratory rate results for normal breathing+rapid breathing and normal breathing+slow breathing based on estimating the number of breaths were better,but the results of the other two cases were not ideal.Conclusion: Based on current data and research,it is the optimal using expiratory and inspiratory phases united estimation method for estimating the respiratory flow rate using tracheal sound.On this basis,Log Variance?100ms,50%?can be used for respiratory flow rate estimation with minimum estimation error,and Log Envelope can be used for respiratory flow rate estimation with optimal performance?considering both estimation error and computational cost?.For the routine monitoring of respiratory rate in clinic,AMDF₁ method is used to monitor respiratory rate based on features of tracheal sound.For respiratory rate monitoring of abnormal respiratory condition?such as intensive care status of respiratory instability?,the method of estimating the number of breaths is used to calculate respiratory rate based on features of tracheal sound to monitor respiratory function.