Akademik Veri Yönetim Sistemi
4th KUWAIT INTERNATIONAL CONFERENCE ON LIFE SCIENCES 2025, Al-Kuwayt, Kuveyt, 11 - 12 Kasım 2025, ss.1, (Özet Bildiri)
Abstract
Tongue movement plays a critical role in the oral stage of swallowing, particularly in boluspreparation, transport, and propulsion. Modern models of oral ingestion describe distinct stages,emphasizing the importance of voluntary tongue activity before swallow reflex initiation.Biomechanically, the tongue facilitates bolus manipulation and transit through coordinatedmovements, supported by synchronized jaw and hyoid actions. Functionally, it acts both as alingual delivery pump, with the anterior tongue functioning like a piston, and as aglossopharyngeal propulsion pump driven by the tongue base. Understanding these phases isessential for characterizing both normal and disordered swallowing.In this study, cortical activity during induced tongue protrusion was examined via EEG, and a deeplearning framework was proposed to classify tongue motor imagery from rest using featuresextracted from specific EEG frequency bands. EEG data were recorded from 30 healthy adultsduring 15 trials comprising alternating rest and imagination tasks. Sixteen electrodes (10–20system) captured data segmented into rest (0–2 s) and protrusion imagination (2.75–4.75 s)intervals. Common Spatial Pattern (CSP) analysis enhanced condition discriminability. PowerSpectral Densities (PSD) were computed across five frequency bands (Delta, Theta, Alpha, Beta,Gamma) using Welch’s method, and average band-specific topographic maps were generated.Topographic EEG images were classified using a custom convolutional neural network (CNN)with leave-one-subject-out cross-validation. The Delta band yielded perfect classificationperformance (Accuracy, F1-score, AUC = 1.000 ± 0.000), while the Theta band also showed strongresults (Accuracy = 0.960 ± 0.041, AUC = 0.989 ± 0.016). Higher frequency bands exhibited lowerperformance (e.g., Gamma: Accuracy = 0.546 ± 0.065, AUC = 0.569 ± 0.074), consistent withmotor-related activity being predominantly represented in lower frequencies.These findings demonstrate that tongue protrusion can be robustly identified using EEG-basedfrequency-topographic representations, particularly in lower frequency bands. This approachoffers a novel, non-invasive method for assessing motor tasks related to swallowing, with potentialapplications in dysphagia rehabilitation and neuromotor evaluation.