Akademik Veri Yönetim Sistemi
Biomedical Signal Processing and Control, cilt.119, 2026 (SCI-Expanded, Scopus)
In this study, a multi-scale spatial attention network (MS-SAN) architecture is proposed to distinguish motor imagery and rest paradigms in electroencephalogram (EEG)-based swallowing analyses. Motor imagery is a mental process associated with motor control and cognitive processes, and EEG is a powerful tool that offers high temporal resolution and accuracy in monitoring these processes. The study investigates the MS-SAN model, which can effectively learn frequency bands in EEG signals and model inter-individual variations. The study employed two distinct experimental paradigms to investigate the processes involved in saliva swallowing and water-induced swallowing. The first paradigm focused on examining the cognitive and motor mechanisms underlying saliva swallowing, while the second paradigm extended this investigation to include the effects of water intake on the swallowing process. This dual approach provided valuable insights into how external factors, such as the presence of water, influence the dynamics of swallowing behavior. CSP (Common Spatial Pattern) filtration was applied to extract spatial patterns from EEG data, and the performance of motor imagery and rest paradigms in different frequency bands (delta, theta, alpha, beta, gamma) were evaluated. For each frequency band, motor imagery and rest states were transformed into topographical representations using spatio-spectral analysis. Statistical analysis, including the Friedman test and post-hoc pairwise comparisons using the Dunn-Sidak test, was conducted to evaluate the differences among EEG frequency bands. The study shows that the MS-SAN model exhibits high classification performance, especially in low-frequency bands (delta and theta), and accurately distinguishes neurophysiological differences between motor imagery and rest states. The results show that the MS-SAN model provides an effective classification method for EEG-based swallowing analysis. These techniques hold great potential for clinical applications, motor rehabilitation, and neurophysiological analysis.