Continuous blood pressure (BP) estimation helps with the hypertension treatment. However, continuous BP estimation is only possible with an invasive catheter measurement method as the gold standard. In this study, we proposed an Attention-U-Net neural network for noninvasive continuous BP estimation using photoplethysmography (PPG) signals. Specifically, the proposed Attention U-Net architecture was evaluated on Physionet’s Cuff-Less Blood Pressure Estimation Dataset. A self-attention mechanism is added in front of each decoder in U-Net, and a feature extraction-recovery layer is added after the last decoder. The experimental results validate the feasibility of the proposed method for continuous BP estimation based on PPG. Meanwhile, the results also show that Attention-U-Net has better performance than Res-Net, Dense-Net, GRU based Seq2Seq and etc. Therefore, the proposed method is a promising alternative for noninvasive continuous BP estimation.
|