Related Experiment Video
Updated: Jul 4, 2026

BrainBeats as an Open-Source EEGLAB Plugin to Jointly Analyze EEG and Cardiovascular Signals
Published on: April 26, 2024
Dynamic beat-to-beat blood pressure estimation using a multi-modal wearable deep learning approach
Qiao Li1, Zichao Shen2, Mohamed Almadi3,4
1Department of Biomedical Informatics, Emory University, Atlanta, GA 30322 United States of America.
This study introduces a novel deep learning method using multi-modal sensor fusion for accurate, calibration-resilient cuffless blood pressure (BP) monitoring. The approach significantly improves BP estimation accuracy during dynamic changes.
Area of Science:
- Biomedical Engineering
- Physiological Monitoring
- Machine Learning in Healthcare
Background:
- Cuffless blood pressure (BP) monitoring faces calibration drift due to reliance on indirect surrogates.
- Existing methods like pulse transit time (PTT) and photoplethysmography (PPG) lack direct pressure measurement.
- Superficial temporal artery tonometry (STAT) offers high-fidelity pressure waveform data but requires integration.
Purpose of the Study:
- To develop a multi-modal deep learning framework for calibration-resilient cuffless blood pressure monitoring.
- To integrate superficial temporal artery tonometry (STAT) with electrocardiography (ECG) and photoplethysmography (PPG) signals.
- To overcome the physiological limitations of current BP monitoring technologies.
Main Methods:
- A custom wearable device simultaneously acquired ECG, PPG, and STAT signals during dynamic BP perturbations.
- Features extracted included heart rate (HR), PTT, and BP-related metrics from STAT.
- A temporal convolutional network (TCN) model was employed to analyze complex, non-linear dependencies between multi-modal features and beat-to-beat BP.
Main Results:
- The TCN model achieved a Mean Absolute Difference (MAD) of 5.58 mmHg for systolic BP, 4.39 mmHg for mean BP, and 4.34 mmHg for diastolic BP.
- Leave-One-Subject-Out Cross-Validation on 29 recordings demonstrated superior performance over PTT and STAT-only baselines.
- The TCN model showed significantly lower errors during dynamic BP fluctuations compared to baseline models (p < 0.05).
Conclusions:
- Fusing tonometry-derived pressure morphology with hemodynamic timing features effectively addresses limitations of conventional PTT methods.
- The proposed multi-modal deep learning framework offers a robust solution for continuous, calibration-resilient BP estimation.
- This technology holds promise for improved non-invasive and accurate blood pressure monitoring.
Related Concept Videos
Assessing Blood pressure using a doppler ultrasound
Pre-Procedural Guidelines for Doppler Ultrasound Blood Pressure Assessment:
Preparation of Equipment:
Equipments Used To Measure Blood Pressure
This invasive approach involves cannulating a peripheral artery. During each cardiac contraction, pressure generates mechanical motion within the catheter, transmitted through rigid, fluid-filled tubing to a transducer. This transducer converts mechanical motion into electrical signals displayed as waveforms on a monitor. An automatic flushing system prevents blood backflow. Due to the potential risk of unexpected arterial blood loss, this method is primarily used in intensive...
Measurement of Blood Pressure
Pre-Procedural Guidelines for Assessing Blood Pressure
Assessment of blood pressure in brachial artery(two-step method)
Neural Regulation of Blood Pressure
Baroreceptor Reflex
Baroreceptors, located in the carotid sinuses and aortic arch, detect changes in blood pressure. When blood pressure rises, these stretch-sensitive receptors...
