Research On Cuff Free Blood Pressure Detection Technology Based On PPG

Prevention and control of hypertension is essential to reduce the risk of cardiovascular diseases.Hypertension,as a chronic disease,is mostly difficult to detect abnormalities,but long-term abnormality is likely to cause the occurrence and development of cardiovascular or renal diseases,therefore daily monitoring of blood pressure is necessary.However,most of the devices used for blood pressure measurement on the market now use the method of cuff inflation and deflation,which will not only bring discomfort to the subject,but also make it difficult to achieve portable measurement anytime and anywhere.For this reason,we propose a more pleasant and portable solution,that is,to measure blood pressure using a single photoplethysmography signal.The specific solution includes signal screening,model building,result analysis,and personalized calibration.First of all,in the model training of PPG signal to predict blood pressure,the quality of the input data source PPG signal is critical.Therefore,an adaptive data filtering method is introduced in addition to the conventional data cleaning method.This method automatically filters part of the abnormal data segments using the statistical distribution of features extracted from the PPG signal(such as heart rate,pulse rise time,rise rate,etc.).Secondly,because in the continuous ABP of aortic blood pressure,we pay more attention to the two key blood pressure values,systolic blood pressure and diastolic blood pressure,we have constructed an asymmetric U-Net network model.This model directly regresses and predicts systolic blood pressure and diastolic blood pressure using only one PPG signal,so that the model focuses more on the two key points,systolic blood pressure and diastolic blood pressure.Finally,from the individual specificity of blood pressure,the personalized calibration module is introduced based on the prediction results of the asymmetric UNet network model.In the individualized calibration,the influence of additional individual characteristics can be excluded as much as possible to build the relationship model between personal PPG signal and blood pressure.