Identifying Key Factors to Improve Autism Spectrum Disorder Diagnosis with Machine Learning

Overview

Autism Spectrum Disorder (ASD) is a neurological disorder characterized by significant social, communication, and behavioral challenges. Current diagnostic practices typically involve a multi- step process, beginning with an initial screening followed by as sessments such as the Social Responsiveness Scale (SRS) to evaluate the severity of ASD. Challenges with the current approach include diagnostic delay, potential inaccuracies, and higher costs, which can delay diagnosis and treatment. Machine learning has the potential to streamline the diagnostic process by employing statistical algorithms to analyze and predict patterns in ASD patient data. The Autism Brain Imaging Data Exchange (ABIDE) is a collaborative database that compiles autistic and allistic brain images and patient data from numerous scientific and educational institutions. The data from ABIDE includes: sex, handedness, age at brain scan, SRS total raw score, SRS raw subscores, SRS adjusted score, SRS adjusted subscores, and SRS edition. SRS subscores include mannerisms, awareness, communication, cognition, and motivation. ABIDE patient data was used to train 𝑘 Nearest Neighbor machine learning algorithms to predict ASD. Results showed that not only can SRS predict ASD, but also that subsets of the total SRS score can be omitted while maintaining high accuracies in ASD diagnosis predictions.

Competition history

• JSHS 2025 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.