Wesper Labホーム睡眠時無呼吸検査の自動睡眠時無呼吸採点アルゴリズムの評価

This study evaluated the performance of the Wesper Lab home sleep apnea test (HSAT) artificial intelligence (AI) automated scoring algorithm under both in-laboratory and real-world conditions. We conducted a multi-tiered validation using two datasets and three analyses. The primary analysis compared apnea-hypopnea index (AHI) and central apnea index (CAI) from Wesper Lab HSATs with simultaneous polysomnography (PSG) scored by blinded technologists (n = 44). The secondary analysis evaluated blinded scoring of raw Wesper Lab signals from the same 44 patients: first by a single scorer, then by two additional scorers to assess inter-scorer consistency. The tertiary analysis examined clinical HSATs (n = 139) in which algorithm-derived AHI was compared with expert rescoring across 11 independent clinics. Agreement metrics included Pearson correlation, Bland-Altman analysis, and confusion matrices. Primary analysis: the algorithm showed strong correlation with PSG for AHI (r = 0.90) and CAI (r = 0.82) with minimal bias on Bland-Altman analysis. Secondary analysis: the correlation was r = 0.95 with minimal bias. Across three scorers, correlation remained ≥0.93. Tertiary analysis: correlation was r = 0.98 with minimal bias. These findings demonstrate that the Wesper Lab autoscoring algorithm is a reliable tool for obstructive sleep apnea and central apnea event detection, supporting its role as an HSAT platform that enhances accessibility to sleep apnea diagnosis.