Our results show that an unattended mandibular repositioning during sleep system using artificial intelligence analytics can predict oral appliance therapy outcome with a high level of accuracy. The AI system significantly outperforms a reasonable heuristic method. Additionally, the MATRx plus MRDS system is highly accurate in predicting the amount of mandibular protrusion required for an individual to achieve therapeutic success. The data also suggest that acceptable symptom resolution can be achieved even when the therapy is not titrated according to such symptoms but rather to an objective index.
The predictive accuracy achieved by the MATRx plus MRDS (sensitivity and specificity > 0.90; 95% lower confidence limit [LCL] > 0.7) is higher than non-MRDS methods such as use of clinical features and pharyngeal airway imaging. One reason for the limited predictive accuracy of airway imaging is that pharyngeal dilator muscles influence the size and posture of the airway, and the airway is typically imaged while the patient is awake. Thus, differences between pharyngeal muscle activity while awake and asleep compromise the predictive accuracy of wakeful airway imaging in predicting outcome during sleep. Another shortcoming of the imaging approach is that pharyngeal collapsibility and sleep-related systemic factors determine the severity of OSA and the response to mandibular protrusion.
Commentaires