Enhancing ABP estimation through comprehensive PPG signal analysis and advanced loss function optimization

The estimation of arterial blood pressure (ABP) using photoplethysmography (PPG) signals has gained significant attention in recent years due to its non-invasive nature and potential for continuous monitoring. Deep learning (DL) techniques have emerged as promising tools for this task, by exploiting the complex relationship between PPG and ABP signals. This study investigates an inter-subject approach for ABP estimation from PPG signal using DL models. The focus of our approach is the design and evaluation of a novel loss function to enhance the accuracy and robustness of ABP estimation across different individuals. We propose the Peak Enhancing Loss Function (PELF) combining mean squared error (MSE) and physiological metrics, which effectively captures both the waveform similarity and clinical relevance of the predicted ABP signal by focusing on the systolic and diastolic points. Through extensive experimentation on different datasets, our results demonstrate that PELF improves the accuracy of ABP estimation compared to conventional ones. In conclusion, this work demonstrates the crucial role of loss function design in optimizing DL models for ABP estimation from PPG signals, advancing the state-of-the-art in non-invasive blood pressure (BP) monitoring. The insights gained contribute to enhancing the reliability and applicability of non-invasive BP monitoring systems in clinical practice.