Abstract
When placed in the oral cavity, we almost immediately recognize the physical characteristics of food, such as its compliance, surface roughness, and geometry. Moreover, we can instinctively judge the amount of mastication necessary to break a food down into a bolus ready to be swallowed. The perceptual encoding of food during stages of oral processing is of significant importance in the research of food science, but efforts to reproduce percepts such as ‘firmness,’ ‘smoothness,’ and ‘thickness,’ via the modulation of tribological and rheological properties have been especially evasive. Furthermore, these efforts have overlooked roles of sensory and proprioceptive feedback from the tongue. In this work, we developed predictive computational models that clarify the interplay of subtypes of sensory neural afferents, and their capacity to contribute to the neural encoding of stimulus diameter, contact geometry, and relative position. First, we employed differential equation models that abstract the neural biophysics in generating mechanosensitive currents and spike firing. Second, we built models of afferent population, varying in density, that encode spatial elements of stimuli such as diameter and contact geometry. Moreover, we leveraged machine learning approaches to classify stimulus spatial elements through their elicited afferent population responses. Our efforts aim in the longer-term development of a computational platform to decode stimulus compliance, surface roughness, and lateral motion, via population response profiles of mechanosensitive afferents, as more layers of complexity in terms of stimulus contact mechanics with the simulated tongue become available. Furthermore, the grand aim of this effort is to provide the foundational steps in creating a computational platform, which can decode complex percepts, e.g. firmness, smoothness, and thickness.
