A supervised learning regression method for the analysis of oral sensitivity of healthy individuals and patients with chemosensory loss

Abstract

AbstractThe gustatory, olfactory, and trigeminal systems are anatomically separated. However, they interact cognitively to give rise to oral perception, which can significantly affect health and quality of life. We built a Supervised Learning (SL) regression model that, exploiting participants’ features, was capable of automatically analyzing with high precision the self-ratings of oral sensitivity of healthy participants and patients with chemosensory loss, determining the contribution of its components: gustatory, olfactory, and trigeminal. CatBoost regressor provided predicted values of the self-rated oral sensitivity close to experimental values. Patients showed lower predicted values of oral sensitivity, lower scores for measured taste, spiciness, astringency, and smell sensitivity, higher BMI, and lower levels of well-being. CatBoost regressor defined the impact of the single components of oral perception in the two groups. The trigeminal component was the most significant, though astringency and spiciness provided similar contributions in controls, while astringency was most important in patients. Taste was more important in controls while smell was the least important in both groups. Identifying the significance of the oral perception components and the differences between the two groups provide important information to allow for more targeted examinations supporting both patients and healthcare professionals in clinical practice.