Introduction / Background: Immune effector cell-associated neurotoxicity syndrome (ICANS) is a frequent complication of chimeric antigen receptor (CAR) T-cell therapy, characterized by encephalopathy, seizures, and cerebral edema. Clinical brain MRI is often normal despite acute clinical symptoms. We tested whether quantitative DTI biomarkers (Mean Diffusivity, MD; Radial Diffusivity, RD; Fractional Anisotropy, FA) provide a more sensitive measure of microstructural injury during ICANS. 

Methods: We conducted a retrospective cohort study using the multi-center CARNIVAL clinical neuroimaging database for pediatric patients who received CD19 and/or CD22-CAR T cells to treat hematologic malignancies. ICANS was graded per ASTCT criteria, ranging from grade 0 (no symptoms) to 5 (death). MRIs were included if they were obtained between day 1-30 after CAR T cell infusion (N=37). If a patient had multiple studies in this timeframe, the study coinciding with the highest ICANS grade was selected. Raw diffusion-weighted DICOM scans were processed through an automated MRtrix3 pipeline incorporating denoising, artifact correction, and voxel-wise diffusion tensor reconstruction to derive whole-brain parenchymal scalar maps of MD, RD, and FA, and white matter tract-specific mean FA. 

Results: MD and RD decreased with increasing ICANS severity (p=0.0450 and p=0.0402, respectively), while median parenchymal FA and tract-specific FA showed no linear correlation with ICANS severity. Among patients with mild ICANS (grade 1-2) whose clinical MRIs were read as normal, several exhibited marked sub-radiological diffusion restrictions with parenchymal MD below  (as low as to ) and RD dropping below , alongside marked FA deviations below . When assessing longitudinal trajectories, both clinical severity ( across MD, RD, and FA) and time post-infusion ( for MD,  for RD, and  for FA emerged as independent predictors.

Conclusion: Quantitative DTI may be a useful biomarker in pediatric ICANS that can detect abnormalities not apparent on routine clinical imaging.