Multiscale Computation in Psychoradiology for Developmental Psychiatric Disorders: A Conceptual Framework

Abstract Developmental psychiatric disorders like ADHD and ASD remain difficult to characterize and treat, mainly due to significant clinical heterogeneity and a lack of objective biomarkers. Conventional neuroimaging often fails to distinguish etiology from correlation and lacks the insights for actionable molecular targets, as most rely on cross-sectional correlations. Here, we present a conceptual framework for multiscale computation that proceeds from longitudinal imaging-marker discovery to cross-scale mechanistic hypothesis generation and molecular imaging-assisted evaluation of interventions. Emotional dysregulation as a transdiagnostic dimension and related research in ASD and ADHD as illustrative cases are used to demonstrate the application of this framework. We first demonstrate how longitudinal trajectory algorithms can identify predictive imaging biomarkers, using developmental deviations in the inferior frontal gyrus and insula associated with emotional dysregulation and related symptoms as illustrative examples. We then describe multiscale fusion models (mFusion) that integrate multi-omics data, including genetics, transcriptomics, and PET, to link neuroimaging markers to candidate molecular pathways, illustrated by findings implicating pathways related to excitatory–inhibitory balance and GABAergic signalling in ADHD and ASD. Finally, we highlight the integration of magnetic resonance spectroscopy (MRS) into clinical trials for intervention-based evaluation, using changes in insular GABA associated with clinical improvement in ASD as an example. Together, these approaches provide a conceptual framework for progressing from imaging-marker discovery toward mechanistic and therapeutic translation.