Stacked Prediction of Activity, Regional Features, and Connectivity
SPARC-CNN is the multibranch resting-state fMRI classification model developed for the manuscript An Adaptive Multifeature Resting-State fMRI CNN for ADHD: Methylphenidate Modulation of Multiscale Intrinsic Brain Dynamics (Monteiro et al., 2026). It integrates spatial fALFF/ReHo, static and dynamic functional connectivity, regional signal features, and demographics through stacked learning to distinguish ADHD from controls.
Trained on ADHD-200 (N = 526), the model was independently evaluated on ePOD-MPH to investigate acute and longitudinal methylphenidate-associated shifts in ADHD classification probability.
Associated manuscript:
Monteiro S, van den Berg J, Vansina E, Reneman L, Caan MWA, Schrantee A. (2026). An Adaptive Multifeature Resting-State fMRI CNN for ADHD: Methylphenidate Modulation of Multiscale Intrinsic Brain Dynamics. Manuscript in preparation.
Also seen at:
Conference presentation: Monteiro S, van den Berg J, Vansina E, Reneman L, Caan MWA, Schrantee A. Acute Methylphenidate-Related Brain Shifts in ADHD Classified by a Multibranch-CNN. CLARA Symposium on AI for Brain Health, Paris, France, 13 November 2026. Poster presentation and 5-minute research pitch.