Mandible behaviour interpretation during wakefulness,sleep and sleep‐disordered breathing

The mandible movement (MM) signal provides information on mandible activity. It can be read visually to assess sleep–wake state and respiratory events. This study aimed to assess (1) the training of independent scorers to recognize the signal specificities; (2) intrascorer reproducibility and (3) interscorer variability. MM was collected in the mid‐sagittal plane of the face of 40 patients. The typical MM was extracted and classified into seven distinct pattern classes: active wakefulness (AW), quiet wakefulness or quiet sleep (QW/S), sleep snoring (SS), sleep obstructive events (OAH), sleep mixed apnea (MA), respiratory related arousal (RERA) and sleep central events (CAH). Four scorers were trained; their diagnostic capacities were assessed on two reading sessions. The intra‐ and interscorer agreements were assessed using Cohen's κ. Intrascorer reproducibility for the two sessions ranged from 0.68 [95% confidence interval (CI): 0.59–0.77] to 0.88 (95% CI: 0.82–0.94), while the between‐scorer agreement amounted to 0.68 (95% CI: 0.65–0.71) and 0.74 (95% CI: 0.72–0.77), respectively. The overall accuracy of the scorers was 75.2% (range: 72.4–80.7%). CAH MMs were the most difficult to discern (overall accuracy 65.6%). For the two sessions, the recognition rate of abnormal respiratory events (OAH, CAH, MA and RERA) was excellent: the interscorer mean agreement was 90.7% (Cohen's κ: 0.83; 95% CI: 0.79–0.88). The discrimination of OAH, CAH, MA characteristics was good, with an interscorer agreement of 80.8% (Cohen's κ: 0.65; 95% CI: 0.62–0.68). Visual analysis of isolated MMs can successfully diagnose sleep–wake state, normal and abnormal respiration and recognize the presence of respiratory effort.