The Impact of rs-fMRI Preprocessing on the Quality of Machine Learning Models for Autism Spectrum Disorder Diagnosis.

Researchers tested different ways to process brain scan data for computer programs that help diagnose autism. They found that how the data is prepared makes a big difference in accuracy - some methods worked very well (up to 95.83% accurate) but didn't work as well when tested on different groups of people. This shows we need better, more flexible approaches and balanced groups of participants to make these diagnostic tools more reliable.

This study examined how different data preprocessing methods affect machine learning models designed to diagnose Autism Spectrum Disorder using brain imaging. Researchers tested 108 different preprocessing approaches on brain scan data from 72 participants, then trained 13 different machine learning classifiers. While some preprocessing configurations achieved up to 95.83% accuracy in initial testing, model performance dropped substantially when tested on broader datasets, indicating poor generalization. The findings highlight that preprocessing choices significantly impact diagnostic accuracy and emphasize the need for adaptive preprocessing strategies and gender-balanced datasets to improve reliability of autism classification models.

While showing promise for machine learning-assisted ASD diagnosis, current models lack sufficient reliability for clinical implementation. Findings suggest need for standardized preprocessing protocols, larger diverse datasets, and improved validation methods before these tools can support clinical decision-making.

Small sample size (72 subjects), unclear generalizability across populations, limited information about participant characteristics, and substantial performance drops in validation testing indicate methodological concerns about model robustness and external validity.

Tools for aiding in the diagnosis of Autism Spectrum Disorder (ASD) using machine learning (ML) and resting-state rs-fMRI (rs-fMRI) must encompass different phases such as data collection, preprocessing, feature extraction, model training, and validation. Many studies rely on a single preprocessing pipeline or use preprocessed data, which might not be optimal for the task at hand. This study investigates the impact of rs-fMRI preprocessing on the performance of ML models for ASD diagnosis. Using a subset of the Autism Brain Imaging Data Exchange (ABIDE) dataset, 72 subjects were preprocessed with 108 different configurations, and features were extracted to train 13 ML classifiers.

Results indicate that preprocessing choices significantly influence model accuracy, with the best configurations achieving up to 95.83% accuracy. However, generalization tests on an extended dataset revealed a substantial performance drop, highlighting challenges in model robustness. Findings emphasize the need for adaptive preprocessing strategies and gender-balanced datasets to improve ASD classification reliability.