ASD Neuro Agentic (ANA): An Agentic Framework for Feature Discovery and Cross Site Generalization in Autism fMRI Classification
Classifiers of autism spectrum disorder (ASD) from resting state functional MRI rarely generalize across sites, and many reported accuracies are inflated by data leakage.We show that an agentic artificial intelligence framework, ASD Neuro Agentic (ANA), whose language model agents propose features from the literature and researcher input while fixed tools perform every calculation, can search…
Autism Spectrum Disorder (ASD) classification using resting state functional MRI data often struggles with site generalization, leading to inflated accuracy claims due to data leakage. Researchers have now developed an agentic artificial intelligence framework called ASD Neuro Agentic (ANA) designed to tackle this issue. <source ab74c89b-0f2> ANA employs language model agents that propose features from literature and user input, while fixed tools perform all calculations, ensuring features are auditably sourced and site-independent.
Using ANA, researchers tested 71 different pipelines on 1,028 participants from 17 ABIDE I centers. The framework selected a ComBat harmonized ensemble of tangent space connectivity models derived from six parcellations and 98 public connectivity derivatives. In leave-one-center-out testing, ANA's ensemble achieved an area under the curve of 0.814 (95% CI, 0.773 to 0.841), surpassing previous ASD classification models ASD DiagNet and ASD SAENet.
The predictive connectivity analysis revealed that ASD individuals exhibited reduced predictive connectivity in their brain networks. ANA's approach provides a systematic and efficient method for exploring, evaluating, and identifying machine learning pipelines, a task that becomes increasingly complex as model selection becomes more intricate.
Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
