A Smartphone-integrated Approach to Blood Pressure Estimation: Combining Video PPG and Pre-trained Convolutional Neural Networks

Non-invasive blood pressure estimation using photoplethysmography signals and artificial neural networks represents a promising advancement in cardiovascular health monitoring. The global prevalence of cardiovascular diseases and the demand for more accessible and practical cardiac health monitoring solutions drive this innovative approach. By leveraging smartphone camera technology to capture photoplethysmography signals, which are subsequently processed by neural network-based software, this method demonstrates significant potential. Results showed notable performance in estimating systolic blood pressure, achieving a mean absolute error of 5.65 ± 3.27 mmHg. Compared to related studies, the proposed method exhibited a lower error rate, surpassed only by a specific study by Yuriy Kurylyak, which reported a mean absolute error of 3.80 ± 3.46 mmHg for SBP estimation. However, the prediction of diastolic blood pressure showed a slightly higher error, with a mean absolute error of 7.51 ± 4.65 mmHg. The significance of this research lies in introducing a promising non-invasive approach to blood pressure estimation, potentially facilitating early detection and more efficient monitoring of cardiac conditions. Furthermore, the implementation of a blood pressure classification system and the potential expansion to mobile devices enhance user convenience and flexibility.