Prognostic Value of Early Magnetic Resonance Imaging Lesion Topography in Diffuse Axonal Injury: The KASHI Classification: An Institute-Based Experience

Diffuse axonal injury (DAI) is a frequent and devastating form of traumatic brain injury (TBI) that requires prolonged intensive care and carries high morbidity and mortality. Early magnetic resonance imaging (MRI) lesion topography may offer crucial prognostic information to guide clinical management and family counseling. The objective of this is to evaluate the prognostic value of early MRI-based lesion location in DAI using the novel KASHI Classification and to correlate lesion topography with acute neurological severity, in-hospital mortality, and long-term functional recovery. In this single-center prospective cohort study, 91 patients with DAI (from 734 TBI admissions) underwent standardized MRI (T1, T2, FLAIR, diffusion-weighted imaging, gradient echo ± diffusion tensor imaging) on Day 1 postinjury. Lesions were graded 1–3C according to the KASHI schema based on their anatomical distribution. Clinical severity was assessed by the Glasgow Coma Scale (GCS) at admission, and outcomes included in-hospital mortality, duration of post-traumatic amnesia (PTA), and 6-month functional status measured by Glasgow Outcome Score and Rancho Los Amigos Scale. Higher KASHI grades were strongly associated with worse acute and long-term outcomes. Grade 3C lesions (pontine, medullary, cerebellar peduncle, or intraventricular hemorrhage) corresponded to the lowest mean GCS scores and a 68 % in-hospital mortality rate. Lesions of the thalamus and basal ganglia independently predicted prolonged PTA and poorer 6-month functional recovery. The KASHI Classification demonstrated high interrater reliability and outperformed total lesion volume models in early prognostication. Early MRI lesion topography, as stratified by the KASHI Classification, provides a simple, reliable framework for predicting neurological severity and recovery in DAI. Its application in acute TBI care can enhance prognostic accuracy and inform clinical decision-making. Further multicenter validation and integration with automated imaging analytics are warranted.