To develop an interpretable, data-driven model that captures the clinico-radiological determinants of reversibility in secondary movement disorders (SMDs) and visualizes them as a reproducible “reversibility signature.”

Reversibility in SMDs has long been described anecdotally; metabolic and vascular etiologies are often recoverable, while post-infectious or structural forms may not be. However, a unified & quantitative framework that translates bedside variables into a predictive map of recovery has not been established. Machine-assisted analytics now allow such patterns to be prospectively identified and interpreted.

In this two-year prospective study, 53 consecutive patients presenting with SMDs were systematically enrolled and followed over 6 months. Variables included age, sex, movement phenotype, etiology, MRI findings and ordinal outcome (1 = normal, 2 = better, 3 = unchanged, 4 = worsened, 5 = death). Outcomes 1–2 were classified as reversible. Machine-assisted clustering and cross-mapping across Etiology, Phenotype & Imaging matrices generated an interpretable multidimensional signature of reversibility.

Overall, 38 of 53 patients (71.7%) demonstrated reversibility. Distinct predictive clusters emerged: metabolic etiologies such as uncontrolled diabetes and hyponatremia; vascular etiologies including ischemic stroke and stroke-related chorea; drug-induced and post-infectious causes showed complete recovery, forming a unified high-reversibility core. In contrast, all subacute sclerosing panencephalitis (SSPE) cases (n = 10) remained non-reversible, with MRI showing either no significant abnormality (8/10) or bilateral cortical/subcortical FLAIR hyperintensities (2/10). Machine-assisted modeling ranked etiology > phenotype > imaging as principal determinants of reversibility.

This machine-assisted prospective analysis introduces a novel reversibility signature that converts conventional clinical data into explainable predictive models. The framework operationalizes recovery prediction in secondary movement disorders, setting the stage for AI-integrated prognostic tools and precision decision-support at the bedside.