Abstract:Objective To construct relatively objective, atlas-based multivariate models for predicting early aphasia severity after stroke, using structural magnetic resonance imaging. Methods We analyzed the clinical and imaging data of 46 patients with post-stroke aphasia. The aphasia severity was identified with a Western Aphasia Battery Aphasia Quotient. The assessments of stroke lesions were indicated by the lesion load of both the cortical language areas (Areas-LL) and four white matter tracts (i.e., the superior longitudinal fasciculus, SLF-LL; the inferior frontal occipital fasciculi, IFOF-LL; the inferior longitudinal, ILF-LL; and the uncinate fasciculi, UF-LL) extracted from human brain atlas. Correlation analyses and multiple linear regression analyses were conducted to evaluate the correlations between demographic, stroke- and lesion-related variables and aphasia severity. The predictive models were then established according to the identified significant variables. Finally, the receiver operating characteristic (ROC) curve was utilized to assess the accuracy of the predictive models. Results The variables including Areas-LL, the SLF-LL, and the IFOF-LL were significantly negatively associated with aphasia severity (p < 0.05). In multiple linear regression analyses, these variables accounted for 59.4% of the variance (p < 0.05). The ROC curve analyses yielded the validated area under the curve (AUC) 0.84 both for Areas-LL and SLF-LL and 0.76 for IFOF-LL, indicating good predictive performance (p < 0.01). Adding the combination of SLF-LL and IFOF-LL to this model increased the explained variance to 62.6% and the AUC to 0.92. Conclusions The application of atlas-based multimodal lesion assessment may help predict the aphasia severity after stroke, which needs to be further validated and generalized for the prediction of more outcome measures in populations with various brain injuries.

Prediction Models Of The Aphasia Severity After Stroke By Lesion Load Of Cortical Language Areas and White Matter Tracts: An Atlas-based Study.