A Review of Artificial Intelligence-Based Dyslexia Detection Techniques

Area of Science:

• Neuroscience
• Computer Science
• Developmental Psychology

Background:

• Dyslexia, a complex learning disorder, is often underdiagnosed due to limitations in traditional assessment methods.
• Artificial intelligence (AI) offers potential for dyslexia detection (DD) using behavioral and neuroimaging data, but faces challenges with limited datasets and model interpretability.
• Dimensionality Reduction Techniques (DRTs) are vital for extracting salient features to enhance Machine Learning (ML) and Deep Learning (DL) based DD models.

Purpose of the Study:

• To review the role of DRTs in improving ML- and DL-based dyslexia detection models.
• To identify critical dyslexia patterns and features extracted by various DRTs.
• To outline current challenges and limitations in the application of DRTs for dyslexia identification.

Main Methods:

• A comprehensive literature search was conducted across major scientific databases (Scopus, Web of Science, PubMed, IEEEXplore).
• 479 articles related to DRTs and dyslexia identification were initially retrieved.
• A rigorous screening process resulted in the inclusion of 39 articles for this review.

Main Results:

• Various DRTs were identified for extracting dyslexia patterns from multimodal data.
• Principal Component Analysis (PCA) was frequently employed for feature extraction and selection in DD studies.
• Key features associated with dyslexia were presented, alongside an analysis of existing DRT challenges.

Conclusions:

• Novel DRTs are needed to enhance the performance and interpretability of AI-driven dyslexia detection.
• Seamless integration of advanced DRTs with Deep Learning (DL) techniques is essential for robust dyslexia identification.
• Further research into DRTs will facilitate more accurate and timely diagnosis of dyslexia.