Discrete wavelet transform-driven optimized deep learning-based framework for dyslexia detection using EEG signals.

Area of Science:

• Neuroscience
• Developmental Psychology
• Machine Learning

Background:

• Dyslexia is a common neurodevelopmental disorder affecting reading and language skills.
• Early identification of dyslexia is crucial for effective intervention and support.
• Current diagnostic methods can be time-consuming and may lack objective measures.

Purpose of the Study:

• To develop an efficient electroencephalography (EEG)-based pipeline for dyslexia detection.
• To utilize deep learning techniques for enhanced accuracy in identifying dyslexia.
• To establish a consistent evaluation protocol for comparing different dyslexia detection models.

Main Methods:

• EEG data were collected from children aged 5-10 during cognitive tasks.
• Discrete Wavelet Transform (DWT) decomposed EEG signals into frequency bands.
• Filter-based feature selection and various machine learning models, including deep neural networks, were employed.
• Performance was benchmarked against classical ML, raw-EEG deep learning, and existing pipelines.

Main Results:

• A compact deep neural network achieved a classification accuracy of 98.85%.
• The proposed DWT-driven deep learning approach significantly outperformed all baseline and re-implemented models.
• The study identified key discriminative neural patterns for dyslexia detection.

Conclusions:

• DWT-based EEG analysis combined with deep learning offers a feasible and accurate method for early dyslexia detection.
• The developed pipeline shows promise as a non-invasive screening tool.
• Early diagnosis through this method can lead to improved educational outcomes via targeted interventions.