Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Clinical neuroimaging as a predictor of recovery from aphasia over time
Poster Session C, Friday, October 25, 4:30 - 6:00 pm, Great Hall 3 and 4
Lily Walljasper1, Deborah F. Levy1, Marianne Casilio1, Sarah M. Schneck1, Jillian L. Entrup1, Anna V. Kasdan1, Michael de Riesthal1, Stephen M. Wilson1,2; 1Vanderbilt University Medical Center, 2University of Queensland
Although acute MRI/CT is highly predictive of long-term outcomes in aphasia after stroke [1,2], the predictive power of clinical neuroimaging for projecting recovery (change in behavior over time) is substantially harder to demonstrate. The subacute-to-chronic period—a critical time frame where acute medical factors have resolved yet substantial recovery is still occuring—poses a particular challenge, given that behavioral change in this timeframe depends strongly on initial aphasia severity. This study aimed to test whether clinical neuroimaging is a meaningful predictor of change in aphasia severity between 1 and 12 months post-stroke onset. We recruited 354 acute ischemic or hemorrhagic stroke patients, of whom 218 presented with aphasia acutely. Patients were evaluated with the Quick Aphasia Battery (QAB), then those with aphasia were followed up at 1 month, 3 months, and 12 months, where possible. The present analysis is based on 47 patients who were tested at the 1 month timepoint, who still presented with aphasia at that time (QAB overall <8.9 out of 10), and were subsequently evaluated at the 12 month timepoint. We used fixed effects linear models to predict aphasia severity at 12 months (QAB overall) based on (1) a baseline model of aphasia severity at 1 month (QAB overall); (2) damage to temporal, prefrontal, and fronto-parietal ROIs (defined using fMRI in a separate group of neurologically normal participants) [3], and total lesion extent, in addition to the baseline; (3) demographic factors (age, sex, handedness, years of education) and stroke type, in addition to the baseline. Prediction r2 and mean absolute error were assessed in leave-one-out cross-validation. In the baseline model, aphasia severity at 12 months was predicted quite well by aphasia severity at 1 month (model r2=62.6%; prediction r2=57.1%; mean absolute error=0.90 QAB scale points). Critically, we found that prediction was substantially and significantly improved by adding neural factors (damage to the 3 ROIs, and total lesion extent) to the model (model r2=79.8%; prediction r2=70.1%; mean absolute error=0.75 QAB points; p<.001 relative to the baseline model). Temporal damage in particular was predictive of poorer recovery: an extensive temporal lesion could reduce the predicted outcome by >3 QAB points. In contrast, prediction was not improved relative to the baseline model by adding only demographic factors (age, sex, handedness, education) or stroke type, either individually or in combination (all p≥.17). In sum, we found that neuroimaging is a significant and meaningful predictor of recovery from aphasia between 1 month and 12 months, above and beyond the predictive power of aphasia severity at 1 month. An important implication of this finding is that we need to make the relevant imaging information accessible to speech-language pathologists. This could involve (1) training neuroradiologists and neurologists to identify critical features and communicate this information in their reports; (2) training speech pathologists to interpret key aspects of certain brain images; and/or (3) developing automated tools to derive aphasia predictions from neuroimaging data.
Topic Areas: Disorders: Acquired,