Dae Hyeon Kim RESEARCH 한국어

Blood Pressure Estimation based on Real-Time Deep Learning Analysis on BCG

Oral presentation

Venue
KOSOMBE Spring Conference
Author role
First author
Year
2021.05

Authors Dae Hyeon Kim, Hyeonwoo Cho, Keonyoung Ko, Jaehyeok Choi, Youjung Seo, Jaewon Yu, Ryanghee Son, Cheolsoo Park

Lab publication record

Research approach

Recorded two-channel BCG at 250 Hz from a chair back and seat. A CNN, BiGRU and temporal attention process ten-second windows to estimate systolic and diastolic blood pressure.

Example systolic cuff measurements and BCG predictions
Example systolic cuff measurements and BCG predictions
View full-size figure
Example diastolic cuff measurements and BCG predictions
Example diastolic cuff measurements and BCG predictions
View full-size figure

Results

Six participants, with approximately three hours of resting and post-exercise measurements each. Within-person temporal splits of 29:14:57 yielded mean SBP/DBP MAE of 4.70/3.83 mmHg and RMSE of 6.02/4.90 mmHg. The study used overlapping windows and normalization before splitting.

Related project

Blood pressure estimation from two-channel BCG

All publications