TY - GEN
T1 - Dash
T2 - 2025 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2025
AU - Hong, Won Seok
AU - Choi, Seunghun
AU - Hong, Kwon
AU - Lee, Woo Hyun
AU - Choi, Hyun Soo
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025
Y1 - 2025
N2 - Obstructive sleep apnea (OSA) is a prevalent sleep disorder with serious health risks, but its gold-standard diagnosis, polysomnography (PSG), is resource-intensive and limits accessibility. To address these limitations, we present DASH (Deep-learning-based Apnea Screening with AHI estimation), a web-integrated end-to-end system for apnea-hypopnea index (AHI) estimation using only single-lead electrocardiography (ECG). DASH consists of three stages: (1) per-segment apnea detection using a fine-tuned ECG-DualNet model, (2) AHI calculation via event aggregation and time normalization, and (3) severity classification based on AHI thresholds. Evaluated on a large-scale PSG-labeled dataset from Kangwon National University Hospital (KNUH), our model demonstrated robust performance, achieving an AUROC of 0.931 (segment-level) and an F1 score of 0.926 (severe-stage). In addition, we developed an interactive web interface for real-time screening and visualization. The results demonstrate that DASH offers a scalable, accurate, and accessible alternative to PSG for OSA detection using only ECG data.
AB - Obstructive sleep apnea (OSA) is a prevalent sleep disorder with serious health risks, but its gold-standard diagnosis, polysomnography (PSG), is resource-intensive and limits accessibility. To address these limitations, we present DASH (Deep-learning-based Apnea Screening with AHI estimation), a web-integrated end-to-end system for apnea-hypopnea index (AHI) estimation using only single-lead electrocardiography (ECG). DASH consists of three stages: (1) per-segment apnea detection using a fine-tuned ECG-DualNet model, (2) AHI calculation via event aggregation and time normalization, and (3) severity classification based on AHI thresholds. Evaluated on a large-scale PSG-labeled dataset from Kangwon National University Hospital (KNUH), our model demonstrated robust performance, achieving an AUROC of 0.931 (segment-level) and an F1 score of 0.926 (severe-stage). In addition, we developed an interactive web interface for real-time screening and visualization. The results demonstrate that DASH offers a scalable, accurate, and accessible alternative to PSG for OSA detection using only ECG data.
KW - AHI
KW - ECG-DualNet
KW - Electrocardiography (ECG)
KW - Obstructive Sleep Apnea (OSA)
UR - https://www.scopus.com/pages/publications/105033563576
U2 - 10.1109/BIBM66473.2025.11356793
DO - 10.1109/BIBM66473.2025.11356793
M3 - Conference contribution
AN - SCOPUS:105033563576
T3 - Proceedings - 2025 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2025
SP - 3668
EP - 3671
BT - Proceedings - 2025 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2025
A2 - Liu, Juan
A2 - Huang, Jingshan
A2 - Wang, Xiaowo
A2 - Zhang, Fa
A2 - Zou, Xiufen
A2 - Tian, Tian
A2 - Hu, Xiaohua
A2 - Hu, Bin
A2 - Xiong, Yi
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 15 December 2025 through 18 December 2025
ER -