Dyslexia Identification Through EEG Signals: Machine Learning Approach

This research investigates the integration of eye tracking and machine learning algorithms for dyslexia detection, achieving high accuracy without relying on conventional classification techniques. Medical professionals are utilizing this advanced method to improve individuals' lives by recognizing challenges associated with dyslexia. We present an innovative supervised learning classifier that utilizes eye-tracking data from 98 individuals diagnosed with dyslexia and 88 who are not. The analysis incorporates a short-time Fourier transform for frequency spectrum evaluation and applies principal component analysis (PCA) for dimensionality reduction. Traditional dyslexia assessments usually encompass reviewing reading and language abilities along with cognitive and psychological dimensions through experimental methods and observational studies. Electroencephalography (EEG) signals play a vital role in brain-computer interfaces; however, predicting dyslexia presents significant challenges. In this study, we analyze attention and mediation parameters using brain signal data from the Massive Open Online Courses (MOOC) dataset, focusing on power spectrum features while implementing binary classification. The aim is to create a structure for predicting dyslexia indicators using EEG analysis, thereby enabling early identification and prompt intervention. A precise diagnosis is essential for identifying dyslexia and excluding other disorders that could affect language skills. To assess the effectiveness of the model, evaluation metrics such as ROC curves and AUC will be employed to gauge precision, recall, and F1-Score. Additionally, methods for feature selection will assist in pinpointing relevant input variables. Moreover, we propose an innovative method that incorporates eye tracking during Q&A sessions alongside facial image feature extraction to enhance understanding of the reading process.