Assessment of Mandibular Movement Monitoring With Machine Learning Analysis for the Diagnosis of Obstructive Sleep Apnea

Abstract

Importance: Given the high prevalence of obstructive sleep apnea (OSA), there is a need for simpler and automated diagnostic approaches.

Objective: To evaluate whether mandibular movement (MM) monitoring during sleep coupled with an automated analysis by machine learning is appropriate for OSA diagnosis.

Design, setting, and participants: Diagnostic study of adults undergoing overnight in-laboratory polysomnography (PSG) as the reference method compared with simultaneous MM monitoring at a sleep clinic in an academic institution (Sleep Laboratory, Centre Hospitalier Universitaire Université Catholique de Louvain Namur Site Sainte-Elisabeth, Namur, Belgium). Patients with suspected OSA were enrolled from July 5, 2017, to October 31, 2018.

Main outcomes and measures: Obstructive sleep apnea diagnosis required either evoking signs or symptoms or related medical or psychiatric comorbidities coupled with a PSG-derived respiratory disturbance index (PSG-RDI) of at least 5 events/h. A PSG-RDI of at least 15 events/h satisfied the diagnosis criteria even in the absence of associated symptoms or comorbidities. Patients who did not meet these criteria were classified as not having OSA. Agreement analysis and diagnostic performance were assessed by Bland-Altman plot comparing PSG-RDI and the Sunrise system RDI (Sr-RDI) with diagnosis threshold optimization via receiver operating characteristic curves, allowing for evaluation of the device sensitivity and specificity in detecting OSA at 5 events/h and 15 events/h.

Results: Among 376 consecutive adults with suspected OSA, the mean (SD) age was 49.7 (13.2) years, the mean (SD) body mass index was 31.0 (7.1), and 207 (55.1%) were men. Reliable agreement was found between PSG-RDI and Sr-RDI in patients without OSA (n = 46; mean difference, 1.31; 95% CI, -1.05 to 3.66 events/h) and in patients with OSA with a PSG-RDI of at least 5 events/h with symptoms (n = 107; mean difference, -0.69; 95% CI, -3.77 to 2.38 events/h). An Sr-RDI underestimation of -11.74 (95% CI, -20.83 to -2.67) events/h in patients with OSA with a PSG-RDI of at least 15 events/h was detected and corrected by optimization of the Sunrise system diagnostic threshold. The Sr-RDI showed diagnostic capability, with areas under the receiver operating characteristic curve of 0.95 (95% CI, 0.92-0.96) and 0.93 (95% CI, 0.90-0.93) for corresponding PSG-RDIs of 5 events/h and 15 events/h, respectively. At the 2 optimal cutoffs of 7.63 events/h and 12.65 events/h, Sr-RDI had accuracy of 0.92 (95% CI, 0.90-0.94) and 0.88 (95% CI, 0.86-0.90) as well as posttest probabilities of 0.99 (95% CI, 0.99-0.99) and 0.89 (95% CI, 0.88-0.91) at PSG-RDIs of at least 5 events/h and at least 15 events/h, respectively, corresponding to positive likelihood ratios of 14.86 (95% CI, 9.86-30.12) and 5.63 (95% CI, 4.92-7.27), respectively.

Conclusions and relevance: Automatic analysis of MM patterns provided reliable performance in RDI calculation. The use of this index in OSA diagnosis appears to be promising.

Conflict of interest statement

Conflict of Interest Disclosures: Dr Pépin reported being a scientific advisor to Sunrise; receiving grants and/or personal fees from ResMed, Philips, Fisher & Paykel, Sefam, AstraZeneca, AGIR à dom, Elevie, VitalAire, Boehringer Ingelheim, Jazz Pharmaceuticals, and Itamar Medical Ltd; and receiving research support for clinical studies from Mutualia and Air Liquide Foundation. Dr Letesson and Messrs Dedave and Denison reported receiving personal fees from Sunrise. Dr Martinot reported being a nonremunerated scientific advisor to Sunrise and being a remunerated investigator in pharmacy trials for Jazz Pharmaceuticals and Theranexus. No other disclosures were reported.

Figures

Figure 1.. Flow Diagram of the Study Protocol

Shown is a comparison between polysomnography (PSG) and automated mandibular movement (MM) analysis procedures. Data concomitantly recorded by in-laboratory PSG and the Sunrise system device were analyzed independently. A, The PSG data were manually scored to export a respiratory disturbance index (PSG-RDI) as the reference method for obstructive sleep apnea (OSA) diagnosis. B, The Sunrise system (Sr) data were automatically uploaded into a cloud-based platform without human intervention, where data were handled by a proprietary machine learning algorithm. After algorithm processing, Sr-RDI was automatically derived for agreement analysis and evaluation of diagnosis performance. ArI indicates arousal index; TST, total sleep time.

Figure 2.. Evaluation of the Agreement Between the 2 Methods of Respiratory Disturbance Index (RDI) Measurement for Obstructive Sleep Apnea (OSA) Diagnosis

The reference method of overnight in-laboratory polysomnography (PSG) is shown on the x-axis. A, Kernel density estimation plot shows the distribution of PSG-derived RDI (PSG-RDI) (discontinued trace) vs the Sunrise system RDI (Sr-RDI) (continuous trace) in the 3 clinical groups. B, Conventional Bland-Altman plot shows the disagreement between PSG-RDI and Sr-RDI (y-axis) as a function of PSG-RDI (x-axis), with individual cases stratified into 3 clinical groups. The horizontal lines indicate the mean difference in the whole sample and within each group. The 2 dashed lines indicate the lower and upper levels (mean, ±1.96 SD) of the disagreement in the whole sample. Bidimensional kernel density estimation plots are superposed to show the distribution of the disagreement as a function of PSG-RDI. The distribution of the disagreement between the 2 methods, stratified by group, is shown on the right.

Figure 3.. Receiver Operating Characteristic Curve Analysis for Evaluating the Performance of the Sunrise System Respiratory Disturbance Index (Sr-RDI) in Obstructive Sleep Apnea Diagnosis

Shown are curves of the binary classification rules to detect patients with obstructive sleep apnea with polysomnography-derived respiratory disturbance index (PSG-RDI) of at least 5 events/h (A) and at least 15 events/h (B) using Sr-RDI. The 95% CIs of the area under the curve (AUC) and smoothing effect were obtained by bootstrapping. The diagonal dotted line serves as a reference and shows the performance if obstructive sleep apnea detection was made randomly.