Abstract 72: A Novel VLSM-CST Lesion Load Model is a Superior Predictor of Motor Outcomes of Acute Stroke Patients
Introduction: Lesion load of the Corticospinal Tract (CST-LL) can predict 64% of the variance in 3-months outcome of acute stroke patients. Voxel-based lesion symptom mapping (VLSM) studies have revealed brain voxels associated with motor impairment. A combined VLSM- CST-LL approach may give particular weight to voxels that are both part of an impairment map and the descending motor tracts.
Hypothesis: A combined VLSM-wCST-LL model can predict acute motor outcome better than weighted CST-LL alone.
Methods: We derived the VLSM map from a group of 50 chronic patients with variable motor deficits relating voxels of patients’ lesions to Upper Extremity Fugl-Meyer (UE-FM) scores. A correction for multiple comparisons was applied at FDR<0.05. Resulting VLSM T-maps were multiplied using our probabilistic CST maps, and then summed to form a canonical VLSM-weighted CST tract. Individual lesion maps from 76 acute stroke patients were overlaid onto the VLSM-weighted CST map to calculate lesion load. Patients were assessed for motor impairment (UE-FM) at baseline and at 3 months. Linear regressions were fit for baseline UE-FM, wCST-LL, and Days-of-Therapy (DoT) to determine predictions of 3-months outcome. A multivariate regression was run using VSLM-weighted CST-LL, controlling for baseline UE-FM and DoT. Age was not a significant regressor. Akaike Information Criterion was run to select the best fit model.
Results: The VLSM analysis determined that voxels of lesions in the precentral gyrus, premotor regions, the corona radiata region, and within the descending motor tracts were significantly related to chronic motor impairment. VLSM-CST-LL applied to a group of acute stroke patients with motor impairment predicted 85% of the variance at 3 months motor outcome. AIC results confirmed with 99% certainty that VLSM is the best fit model.
Conclusions: VLSM-weighted CST-LL is the superior fit model compared to the weighted CST-LL model for predicting 3 months outcome.
Author Disclosures: J. Wang: None. W. Feng: Research Grant; Significant; AHA 14SDG1829003. P.Y. Chhatbar: None. G. Schlaug: Research Grant; Significant; RO1 DC008796; R01 DC009823, Richard and Rosalyn Slifka Family Fund, Tom and Suzanne McManmon Family Fund.
This research has received full or partial funding support from the American Heart Association, National Center.
- © 2016 by American Heart Association, Inc.