Prediction of intradialytic hypotension using PPG signal features

One of the most prevalent complications in hemodialysis patients is repetitive hypotension during dialysis sessions. Different factors can be used for monitoring patient conditions and preventing Intra-Dialytic Hypotension (IDH) occurrence during hemodialysis, such as blood pressure, blood volume, electrical Impedance factors etc. We predicted hypotension episodes by using finger PPG signal features. Because of non-stationary nature of PPG signal, we divided main signal in 5-minute parts with no overlap and analyzed each part, separately. Four different signals in different frequency bands extracted from each part and considered for analyzing in frequency domain. Finally, we extracted 12 features in time domain and 10 features in frequency domain. Using Genetic Algorithm (GA) and “AdaBoost” for feature selection and classification, we diffracted IDH and Pre-IDH episodes of dialysis sessions. The obtained results indicate that the mean value of accuracy, sensitivity and specificity of the proposed algorithm are 90.68%, 86.03% and 93.02% respectively.