Intradialytic hypotension related episodes identification based on the most effective features of photoplethysmography signal.

BACKGROUND AND OBJECTIVE: One of the most adverse conditions facing the hemodialysis patient is repetitive hypotension during their dialysis session. Different factors can be used to monitor patient conditions and prevent Intradialytic Hypotension (IDH) during hemodialysis. These factors include blood pressure, blood volume, and electrical Impedance factors. In this paper, pre-IDH and IDH episodes were recognized and classified by using the features of the finger photoplethysmography (PPG) signal. In other words, the goal of present study is to use PPG signal features to predict the risk of acute hypotension.

METHODS: Since the PPG signal is non-stationary in nature, the main signal was divided in five-minute intervals with no overlap and then each interval was analyzed separately and fifteen PPG signal features in time and seven features in the frequency domain were extracted. Then different feature selection and classification methods were applied on the normalized feature matrix to select the best features and detect IDH and pre-IDH episodes in dialysis sessions.

RESULTS: The best results were achieved from a genetic algorithm and AdaBoost. The obtained results on our developed database indicated that the mean and maximum accuracy of the proposed algorithm were 94.5 ± 1.0 and 96.6 respectively.

CONCLUSION: Some PPG signal features can be useful during hemodialysis session for hypotension management.