DNN-BP: a novel framework for cuffless blood pressure measurement from optimal PPG features using deep learning model

Area of Science:

• Biomedical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Continuous blood pressure (BP) monitoring is crucial for health management but is hindered by uncomfortable cuff-based devices.
• Photoplethysmography (PPG) signals offer a non-invasive alternative for BP monitoring.
• Existing PPG-based BP estimation methods require improvement in accuracy and reliability.

Purpose of the Study:

• To develop and validate a novel algorithm for continuous BP monitoring using only PPG signals and deep neural networks (DNNs).
• To enhance BP estimation accuracy by employing advanced signal processing and an ensemble feature selection technique.

Main Methods:

• PPG signals from 125 subjects (218 records) were acquired and processed to remove noise and artifacts.
• Pulse wave analysis was performed on PPG signals to extract domain-specific features.
• An ensemble feature selection method combined four subsets to identify optimal features for DNN-based BP prediction.

Main Results:

• The proposed DNN algorithm with ensemble feature selection achieved high accuracy in estimating systolic blood pressure (SBP) (R²=0.962, MAE=2.480 mmHg) and diastolic blood pressure (DBP) (R²=0.955, MAE=1.499 mmHg).
• The algorithm met the Advancement of Medical Instrumentation standard for SBP and DBP estimations.
• Results attained Grade A according to the British Hypertension Society standard for both SBP and DBP.

Conclusions:

• Continuous BP can be accurately estimated using PPG signals with an optimal feature set and DNN models.
• The developed algorithm demonstrates significant potential for integration into mobile healthcare devices for convenient BP monitoring.
• This non-invasive approach offers a promising alternative to traditional cuff-based BP measurement devices.