Related Experiment Video
Updated: Aug 5, 2026

Software for Analysis of Heart Rate and Blood Pressure Time-series Data from the Valsalva Maneuver
Published on: June 27, 2025
Practical Optimization of Deep Learning Models for Cuffless Blood Pressure Estimation From Photoplethysmography
Vikas Khurana1, Mohammad Aejaz Pamidi2, Bharat Bhushan Joshi Bhushan Joshi2
1The Wright Center for Graduate Medical Education, 501 S Washington Avenue, Scranton, Pennsylvania, 18505, United States.
Optimizing deep learning for cuffless blood pressure (BP) estimation using photoplethysmography (PPG) requires careful selection of sampling frequency and derivation methods. This study found 62.5 Hz and multi-peak averaging yielded best results, informing device design.
Area of Science:
- Biomedical Engineering
- Signal Processing
- Artificial Intelligence
Background:
- Continuous, non-invasive blood pressure (BP) monitoring is crucial, especially when cuff-based methods are impractical.
- Deep learning models show promise for estimating BP from photoplethysmography (PPG) signals.
- Optimizing model inputs and signal processing is key to improving accuracy.
Purpose of the Study:
- To enhance deep learning-based cuffless BP estimation using PPG.
- To evaluate the impact of sampling frequency, BP derivation methods, and respiratory rate on estimation accuracy.
- To identify error patterns, particularly at BP extremes, for improved device design.
Main Methods:
- Analysis of 205,850 paired PPG-BP segments from the MIMIC database.
- Utilized a convolutional Multi-resolution U-Net (MultiResUNet) model for PPG-to-BP waveform translation.
- Compared input sampling frequencies (125, 62.5, 31.25 Hz) and SBP/DBP derivation techniques (min-max vs. multi-peak averaging).
Main Results:
- 62.5 Hz sampling frequency reduced mean absolute error (MAE) compared to 125 Hz (SBP: 5.24 vs. 6.34 mmHg; DBP: 2.87 vs. 3.27 mmHg).
- 31.25 Hz significantly increased MAE (SBP: 16.09 mmHg; DBP: 9.10 mmHg).
- Multi-peak averaging outperformed minimum-maximum extraction; higher errors observed at BP extremes and certain respiratory rates.
Conclusions:
- Optimal sampling frequency (62.5 Hz) and derivation method (multi-peak averaging) improve cuffless BP estimation accuracy.
- Understanding error patterns related to respiratory rate and BP extremes is vital for clinical applicability.
- Further validation on subject-level datasets is necessary to meet clinical standards for PPG-based BP monitoring devices.
Related Concept Videos
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...
Pre-Procedural Guidelines for Assessing Blood Pressure
Assessing Blood pressure using a doppler ultrasound
Pre-Procedural Guidelines for Doppler Ultrasound Blood Pressure Assessment:
Preparation of Equipment:
Sites for measuring blood pressure
The Brachial Artery: Primary Site for Blood Pressure Measurement
Special considerations while measuring blood pressure
Monitoring Both Arms:
Monitoring BP in both arms during the initial assessment is advisable, as the systolic value may differ by five to ten mm Hg between arms. For subsequent BP assessments, use the arm with the higher reading.