A narrative review of quantitative MRI and radiomic biomarkers for prognostication in traumatic brain and spinal cord injury

Abstract Background Traumatic brain injury (TBI) and spinal cord injury (SCI) cause substantial long-term disability. Conventional clinical scales and imaging have limited sensitivity to microstructural and metabolic injury, contributing to prognostic uncertainty. Quantitative MRI (qMRI) and radiomic biomarkers offer objective measures of microstructure, myelin content, hemorrhage burden, and tissue heterogeneity that may improve outcome prediction. Objective To synthesize evidence on qMRI and MRI-based radiomic biomarkers for prognostication in TBI and SCI, critically evaluate methodological quality and validation practices, and outline pathways for clinical translation. Methods We performed a narrative synthesis of systematic reviews, meta-analyses, and primary clinical studies evaluating qMRI modalities (DTI, NODDI, DKI, MTI, MRS, SWI, perfusion) and radiomic approaches in adult TBI and SCI. No meta-analysis was performed due to heterogeneity of imaging protocols and endpoints. Results qMRI modalities, including DTI, DKI, NODDI, MRS, SWI, T1/T2 mapping, myelin-water imaging, and MR perfusion, demonstrate robust correlations with integrity of white matter, axonal injury, neuroinflammation, metabolic dysfunction, microhemorrhage burden, and perfusion derangements. Reduced fractional anisotropy, increased mean diffusivity, increased extracellular volume fraction, and metabolite disturbances are common parameters that predicted functional outcomes across cohorts with TBI and SCI. Several studies achieve high prognostic accuracies using radiomics and machine-learning models incorporating texture, shape, wavelet, and deep-learning features, often superior to conventional imaging. Conclusion qMRI and radiomics hold promise for objective prognostication in neurotrauma but require harmonized acquisition, IBSI-compliant pipelines, large multicenter validation, and demonstration of clinical utility before routine adoption.