Photoplethysmography Based Stratification of Blood Pressure Using Multi-Information Fusion Artificial Neural Network

Dingliang Wang, Xuezhi Yang, Xuenan Liu, Shuai Fang, Likun Ma, Longwei Li; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 276-277

Regular monitoring of blood pressure (BP) is an effective way to prevent cardiovascular diseases, especially for elderly people. At present, BP measurement mainly relies on cuff-based devices which are inconvenient for users and may cause discomfort. Therefore, many new approaches have been proposed to achieve cuff-less BP detection in recent years. However, the accuracy of the existing approaches still needs to be improved. In this study, holistic-based PPG and its first and second derivative features are extracted and a new multi information fusion artificial neural network (MIF-ANN) is designed to effectively fuse and exploit multiple input data. Experimental results on a public database which contains 12000 subjects show that the proposed network can model the relation between Photoplethysmography (PPG) and BP well, achieving averagely accuracy of 91.33% for 5-category BP stratification. Additionally, this study verified that multi information fusion based on meticulously designed network plays an important role in improving the accuracy of BP detection.