Autism Data Classification Using AI Algorithms with Rules: Focused Review.

Area of Science:

• Machine Learning
• Developmental Neuroscience
• Clinical Psychology

Background:

• Autism Spectrum Disorder (ASD) screening is challenging due to symptom variability and subtle early signs.
• Machine learning (ML) for ASD faces hurdles with data diversity, symptom management, and model interpretability.
• Interpretable classifiers offer transparency, crucial for clinical trust and adoption in ASD diagnosis.

Purpose of the Study:

• To review recent research on rule-based classification for ASD detection from a behavioral perspective.
• To consolidate current findings, identify research gaps, and guide future studies in interpretable ASD diagnostics.
• To enhance understanding of ML techniques for early ASD detection and intervention.

Main Methods:

• Review of recent literature on rule-based classification algorithms applied to ASD detection.
• Analysis of datasets, model performance, and identified behavioral features.
• Exploration of hybrid AI approaches combining deep learning with rule-based classifiers.

Main Results:

• Rule-based classifiers enhance transparency and understanding of ASD diagnostic models for clinicians.
• Interpretable models facilitate the identification of key behavioral patterns indicative of ASD.
• Hybrid AI approaches show potential for improved accuracy and interpretability in ASD detection.

Conclusions:

• Interpretable classification, particularly rule-based methods, is vital for advancing early ASD detection and intervention.
• Integrating advanced AI with rule-based systems offers a promising path for accurate, transparent ASD diagnostics.
• Further research is needed to consolidate findings and guide the clinical application of these techniques.