Diagnostic Evaluation of Automatic Noncontact Ultra-Wideband Radar Combined With Oximeter for Obstructive Sleep Apnea in the South of China.

Background: Polysomnography (PSG) is the traditional technique for diagnosing obstructive sleep apnea (OSA) with some limitations. Ultra-wideband radar (UWB) is a new method for diagnosing OSA that combines multiple techniques including radar technology, artificial intelligence (AI), and big data algorithms. The accuracy of UWB in OSA diagnosis needs further scientific verification, especially in southern China. Methods: Fifty patients from southern China wore UWB with oximetry and PSG simultaneously overnight. UWB generated automated reports; PSG was manually interpreted. Lowest oxygen saturation (LSpO2) and apnea-hypopnea index (AHI) from both methods were compared. Results: High correlation was found between UWB and PSG for AHI (r = 0.925, p < 0.001) and LSpO2 (r = 0.990, p < 0.001). The average bias of AHI (-3.09, p=0.98, 95% CI -15.48 - 9.31) and LSpO2 (-0.34, p > 0.99, 95% CI -3.68 - 3.04) between the two methods was small. ROC analysis showed good diagnostic performance of UWB versus PSG (AUC = 0.979, p < 0.001), with 95.8% sensitivity and 100% specificity. Conclusion: The study demonstrated UWB combined with oximetry could be a reliable alternative to PSG for diagnosing OSA in the south of China.