UC Irvine

This thesis reports behavioral and neurological results of a novel form of aphasia therapy – based on imitation of audiovisual speech – that was motivated by neurophysiological findings in human and non-human primates. The six-week intensive therapy program was completed by nineteen participants with post-stroke aphasia.

Participants demonstrated significant improvement on the practiced repetition task and generalization to other tasks. Measures included subtests of the Western Aphasia Battery-Revised (Aphasia Quotient, Cortical Quotient, Repetition, Naming and Word Finding) and specific characteristics of narrative production (number and percent correct information units).

This research investigated the role of performance variability within individuals during repetition to predict improvement with practice. We found that individuals demonstrating greater variability at baseline made greater improvement following treatment. Additionally, changes in variability over the course of treatment were negatively correlated with changes in performance. The more participants improve, the more consistent (less variable) their performance.

On the narrative task, significant improvement was positively correlated with the number of therapy sessions that were completed, a result not explained by other traits that are often associated with aphasia recovery, such as lesion size and time since onset. The imitation therapy generalizes to other tasks, and more practice confers greater benefit.

With functional neuroimaging during the resting state, we discovered two inter-related functional predictors of improvement in narrative production. Using a sliding window approach, we investigated the dynamic nature of resting state networks (RSNs) as they change over the course of therapy. An increase in the amount of time spent in one of the states – a state characterized by minimal correlation among the identified RSNs – predicted improvement on the narrative task. We interpreted this finding as evidence for adaptive segregation among the RSNs.

Using a graph theoretical approach, we found a second functional predictor. By grouping the RSNs into communities, we examined network changes representative of segregation in the brain, and found increased RSN modularity to be positively correlated with behavioral improvement.

These findings of previously unexplored behavioral and neurological changes and predictors associated with post-therapy recovery are consistent with an emerging approach to personalized (precision) aphasia treatment.