0579 Feasibility of a Patch-Type Abdominal Sensor for Sleep Apnea and Hypopnea Detection: A Validation Study

Abstract Introduction Polysomnography (PSG) remains the standard for diagnosing sleep apnea but is limited by its high cost and in-lab setting. The ATP-T200 (ATsense, Korea) is a novel patch-type abdominal sensor that measures sleep respiration via capacitance changes reflecting distance between the patch electrode and the abdominal surface. This study evaluated its feasibility for detecting sleep respiratory events (apnea and hypopnea) compared with PSG recordings. Methods Adults (≥19 years) scheduled for PSG for clinical reasons were recruited.. The patch-type sensor was attached to the abdomen simultaneously with standard in-lab PSG. Abdominal signals were pre-processed and initially screened for event candidates using a rule-based algorithm, followed by refined detection with the Abdomen Abnormal Detection Net (AADNet), an AI model that learns local respiratory patterns and temporal dependencies for accurate event classification.Data were divided 1:1 into training and test sets for model development and evaluation. The accuracy of apnea–hypopnea index (AHI) estimation and epoch-by-epoch event detection were assessed, and the area under the receiver operating characteristic curve (AUROC) values were calculated using AHI thresholds of ≥15, and ≥30 events/h. Results After 8 were excluded due to poor signal quality, the final analysis included 88 participants (mean age 48.5 ± 15.5 years, 81% men, BMI 26.0 ± 4.2 kg/m2). Participants were randomly divided into training (n = 44) and test (n = 44) sets, stratified by sex, age, and BMI without significant group differences. For the AHI, strong correlation was observed between the patch-based and PSG measurements (Pearson r = 0.902, p 0.001), and Bland–Altman analysis showed a mean difference of 0.82. When classifying patients by AHI thresholds, AUROC values were 0.963 for both AHI ≥ 15 (29/44; sensitivity 89.2%, specificity 85.7%) and AHI ≥ 30 (15/44; sensitivity 93.3%, specificity 93.1%). Epoch-by-epoch analysis demonstrated sensitivity 83.3%, specificity 91.6%, accuracy 89.5%, and F1-score 80.0. Conclusion The patch-type abdominal sensor showed strong agreement with PSG and accurately estimating AHI across clinically relevant thresholds. These findings support the feasibility of abdominal capacitance–based respiration monitoring combined with AI-driven event detection as a practical approach for screening and follow-up outside the sleep laboratory. Support (if any) This study was supported by ATsense, Korea.