Cuffless Blood Pressure Estimation Model Based on Prior Information of Physiological Data and LSTM

Blood pressure (BP), as a key indicator, reflects the health status of the cardiovascular system. To address the issue of accurate BP measurement, a BP estimation model based on electrocardiogram (ECG) and photoplethysmography (PPG) and prior information on pulse transit time (PTT) has been proposed. This model features a parallel feature extraction module based on long short-term memory (LSTM) for extracting features and incorporates the prior information PTT through a pretraining strategy. The performance of the model is analyzed using the UCI cuffless BP estimation dataset (UCI-BP) along with correlation analysis, Bland–Altman analysis, and statistical histogram analysis of estimation error. The results show that the mean absolute error (MAE) and standard deviation (SD) for systolic BP (SBP) are 1.02 and 1.49 mmHg, respectively, and for diastolic BP (DBP) are 3.16 and 3.74 mmHg, respectively. These results meet the BP standards of the American Association for Medical Devices (AAMI) and the class A standard of the British Hypertension Society (BHS).