Machine learning-based ABA treatment recommendation and personalization for autism spectrum disorder: an exploratory...

Area of Science:

• Neuroscience
• Artificial Intelligence
• Developmental Psychology

Background:

• Autism spectrum disorder (ASD) is a neurodevelopmental condition impacting social communication and behavior.
• Applied Behavior Analysis (ABA) is a primary treatment for ASD, but faces challenges.
• Shortages of ABA practitioners and clinician subjectivity hinder effective, data-driven treatment planning.

Purpose of the Study:

• To develop and evaluate machine learning algorithms for recommending and personalizing ABA treatment goals for individuals with ASD.
• To address limitations in current ABA practice, including practitioner shortages and subjective decision-making.

Main Methods:

• Two machine learning algorithms (patient similarity and collaborative filtering) were applied to personalize ABA treatment goals.
• The models predicted ABA treatment recommendations for 29 participants with ASD.
• Treatment efficacy was assessed by the percentage of mastered goals.

Main Results:

• Machine learning models achieved 81-84% accuracy in predicting ABA treatment goals.
• Normalized Discounted Cumulative Gain (NDCG) scores ranged from 79-81%, indicating strong performance compared to clinician recommendations.
• Treatment efficacy varied, with the percentage of mastered goals indicating the success of recommended interventions.

Conclusions:

• Machine learning offers a viable solution to personalize ABA treatment goals for ASD, overcoming current practice limitations.
• The proposed strategy demonstrates generalizability to other interventions and neurological disorders.
• This data-driven approach can enhance the quality and accessibility of ASD interventions.