Back

Redefining Autism Subtypes: a machine learning approach leveraging topological data analysis, network measures and hemispheric lateralization

Alves, C. L.; Sallum, L. F.; Aguiar, P. M. d. C.; Porto, J. A. M.; Rodrigues, F. A.; Toutain, T. G. L. d. O.; Moeckel, M.

2025-04-16 psychiatry and clinical psychology
10.1101/2025.04.14.25325816 medRxiv
Show abstract

Autism subtypes, including general Autism Spectrum Disorder (ASD) and Asperger Syndrome (AS), exhibit distinct neural connectivity patterns. This study is the first to systematically integrate Topological Data Analysis (TDA) with complex network measures and machine learning (ML) to investigate brain lateralization and connectivity differences among these subtypes. Using fMRI-derived connectivity matrices, TDA metrics--such as persistence entropy and fractal dimension--revealed that AS networks are highly integrated and hierar-chically complex, distinguishing them from both ASD and typically developing (TD) groups. Shapley Additive Explanations (SHAP) analysis identified the left primary motor cortex as a key feature across all subtypes, and highlighted its subtype-specific correlations with other brain regions. ML models trained on these features achieved high classification accuracy, with an AUC of 0.983. This fMRI-based analysis supports the classification of AS as a distinct group alongside ASD due to its unique neurobiological characteristics.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.