A computer-based method to estimate the level of sensitivity of typical somatosensorial responses

Citation
Share
Abstract
Understanding somatosensory responses is fundamental to human interaction with the environment, yet quantitative tools for assessing typical tactile responses remain underdeveloped. This thesis introduces a novel computer-based method to evaluate somatosensory processing through electroencephalographic data, focusing on responses to different tactile stimuli. The project will be conducted in three stages: 1) registration of typical somatosensory evoked responses due to touch, air, and vibration in incremental intensities using electroencephalography, 2) validation of the prototypes to evoke tactile evoked potentials, 3) development and evaluation of a classification model to differentiate tactile stimuli and intensities. The study involved the creation of a database of Electroencephalographic recordings from 34 healthy adult volunteers exposed to air, vibration, and caress stimuli, under four diffrent intensity levels intensity levels. The neural responses were analyzed using Discrete Wavelet Transform and classified with machine learning models including K-Nearest Neighbors, Random Forest, and Multilayer Perceptron. For a generalized classification model, an accuracy of 72.6% was achieved for distinguishing stimulus type, 39.3% accuracy for intensity classification and 33.4% for both stimulus type and intensity. Individual classifiers for each subject had an increase in accuracy of 6-10%. Additionally, a deep learning model, EEGNet, was implemented, yielding similar results for stimulus type but lower performance for intensity. Analysis revealed significant inter-subject variability, with subject-specific models outperforming generalized ones, highlighting the need for individualized approaches in somatosensory assessments. This study offers a novel dataset and model framework, which enhances the understanding of neural tactile processing to advance sensory-based interfaces and diagnostic tools in neurophysiological research.
Description
https://orcid.org/0000-0002-2256-2958