Year: 2026
Type: article
Abstract: Attention Deficit Hyperactivity Disorder (ADHD) is a common neurodevelopmental disorder characterized by heterogeneous clinical manifestations and diagnosis uncertainty. Despite extensive research, a lack of quantitative, reproducible, and anatomically interpretable biomarkers continues to limit objective ADHD characterization from structural MRI. This study explored morphological ADHD-related brain patterns using a Curvelet-based representation applied to structural MRI from 875 ADHD-200 participants, and performed an additional external validation with 79 participants from the independent WMR-ADHD dataset. The proposed framework projects regional brain anatomy into three 81-dimensional feature vectors corresponding to the Generalized Gaussian distribution parameters approximating the subband curvelet coefficients. Unsupervised classification achieved balanced accuracy(BA) of 0.78 and 0.67 (both \(p<0.001\) ) without diagnostic labels, outperforming supervised nested cross-validation (BA=0.56–0.59 and class-balanced F-macro=0.54–0.58), while cross-dataset generalization yielded consistent above-chance discrimination (BA=0.66–0.68). Across experiments, the cerebellum was the region with higher classification scores, aligned with its reported role in ADHD, although region selection varied across acquisition sites. The observed gap between unsupervised and supervised performance suggests that current diagnostic boundaries may not fully drive the neuroanatomical variation captured by Curvelet features, yet these findings support Curvelet-based representations as an interpretable framework for characterizing ADHD-related brain morphology.
Source: Scientific Reports
Institution: Universidad Nacional de Colombia
Language: English
Subfield: Psychiatry and Mental health
Field: Medicine
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $2,690
OpenAlex ID: W7203837002
FWCI:
Cites:
Cited by:
Related to: