338 Diffusion Basis Spectrum Imaging (DBSI) Prognosticates Outcomes for Cervical Spondylotic Myelopathy after Surgery

Abstract

OBJECTIVES/GOALS: Diffusion basis spectrum imaging (DBSI) allows for detailed evaluation of white matter microstructural changes present in cervical spondylotic myelopathy (CSM). Our goal is to utilize multidimensional clinical and quantitative imaging data to characterize disease severity and predict long-term outcomes in CSM patients undergoing surgery. METHODS/STUDY POPULATION: A single-center prospective cohort study enrolled fifty CSM patients who underwent surgical decompression and twenty healthy controls from 2018-2021. All patients underwent diffusion tensor imaging (DTI), DBSI, and complete clinical evaluations at baseline and 2-years follow-up. Primary outcome measures were the modified Japanese Orthopedic Association score (mild [mJOA 15-17], moderate [mJOA 12-14], severe [mJOA 0-11]) and SF-36 Physical and Mental Component Summaries (PCS and MCS). At 2-years follow-up, improvement was assessed via established MCID thresholds. A supervised machine learning classification model was used to predict treatment outcomes. The highest-performing algorithm was a linear support vector machine. Leave-one-out cross-validation was utilized to test model performance. RESULTS/ANTICIPATED RESULTS: A total of 70 patients – 20 controls, 25 mild, and 25 moderate/severe CSM patients – were enrolled. Baseline clinical and DTI/DBSI measures were significantly different between groups. DBSI Axial and Radial Diffusivity were significantly correlated with baseline mJOA and mJOA recovery, respectively (r=-0.33, p<0.01; r=-0.36, p=0.02). When predicting baseline disease severity (mJOA classification), DTI metrics alone performed with 38.7% accuracy (AUC: 72.2), compared to 95.2% accuracy (AUC: 98.9) with DBSI metrics alone. When predicting improvement after surgery (change in mJOA), clinical variables alone performed with 33.3% accuracy (AUC: 0.40). When combining DTI or DBSI parameters with key clinical covariates, model accuracy improved to 66.7% (AUC: 0.65) and 88.1% (AUC: 0.95) accuracy, respectively. DISCUSSION/SIGNIFICANCE: DBSI metrics correlate with baseline disease severity and outcome measures at 2-years follow-up. Our results suggest that DBSI may serve as a valid non-invasive imaging biomarker for CSM disease severity and potential for postoperative improvement.