LAUSR

The oxygen desaturation index (ODI) quantifies the number of desaturations per hour of sleep and is a valuable parameter in diagnosing sleep-related breathing disorders. While oximetry alone is insufficient for an accurate diagnosis, it provides meaningful insights for Obstructive Sleep Apnea (OSA) and the ODI has been shown to correlate with the Apnea-Hypopnea Index (AHI). The close relationship between the ODI and other parameters is especially interesting for home-based sleep studies and/or screening, which typically use a limited set of sensors (but almost always include oximetry) and are becoming increasingly more prevalent. An accurate and automated tool for oximetry analysis will thus be paramount for the sleep clinician. The current study aims to evaluate a novel automated solution for oxygen desaturation scoring, provided by Somnilog. We selected 20 overnight recordings, a mix of hospital- and home-based recordings from a dataset from the Antwerp University hospital and an internal database collected from several French sleep centers. The analyzed population (10 females, age 53.0±16.7 years old, BMI of 28.8±4.8 kg/m²) had a variety of diagnoses, including non-apneic snoring and central and obstructive sleep apnea. Each recording was also scored for oxygen desaturations by a sleep expert, using the AASM guidelines and a 3% drop criterion. The automated solution accurately detected desaturation events with a precision and recall of 0.97 and 0.96, respectively. The reported ODI had an intraclass correlation (ICC) of 0.99 with the index found by the expert. The difference in the number of events reported by the automated solution and the expert was not significant between the lab- and home-recorded patients (p=0.23). We furthermore validated the novel solution with another well-used commercial package and obtained similar performance metrics: precision and recall of 0.96 and ICC of 0.99. The current study demonstrates the validity of the novel commercial package for oxygen desaturation scoring. The automated tool achieved near-perfect agreement with the sleep experts, and a similar performance to another existing solution in the market. The tool is robust for home-based recordings which are typically more artifact-prone, thereby supporting the increasing demand for out-of-centre overnight studies. n/a