You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster E49, Friday, November 10, 10:00 – 11:15 am, Harborview and Loch Raven Ballrooms

Behavior stability and reliability of fMRI activation in stroke aphasia

Brielle Stark1, Grigori Yourganov1, Alexandra Basilakos1, Chris Rorden1, Leonardo Bonilha2, Julius Fridriksson1;1University of South Carolina, 2Medical University of South Carolina

Brain changes associated with aphasia treatment have been addressed in numerous studies over the past two decades. Neural activation measured during multiple pre-testing time-points must be reliable in order to relate neural changes post-treatment to intervention and not to inherent variability. Language tasks during fMRI have been shown to elicit moderately reliable activation in neurotypical adults (Wilson et al 2016). The reliability of language tasks in stroke aphasia has been investigated in small sample sizes with varying reliability measurements. Verifying fMRI reliability in stroke aphasia is important, given the prevalence of using longitudinal fMRI language tasks in assessment of intervention. The purpose of this study was to investigate fMRI activation reliability, taking into account the stability of language impairments, in a relatively large population of stroke survivors with aphasia. Sixty participants (22 F) were recruited at the University of South Carolina and Medical University of South Carolina. Participants had one-time left hemisphere ischemic stroke; were at least six months’ post-stroke (M=37.68 months, SD=46.61); average age-at-testing 60.1 years (SD=10.09); and average lesion volume 140518.23 mm3 (SD=91769.02mm3). Participants were not undergoing speech-language therapy when enrolled in the study. Twenty-five participants had Broca’s aphasia, five had Wernicke’s aphasia, seventeen had anomic aphasia, nine had conduction aphasia, and four had global aphasia. During two baseline fMRI scans conducted one week apart, participants named aloud 40 object pictures and remained silent during 20 abstract pictures. Responses were scored offline as correct or not correct. We tested behavior stability in two ways. To determine the stability of item-level responses for each participant, i.e. whether a correct response at session one was likewise correct at session two, we computed Cohen's Kappa. We computed Pearson's product moment correlation to measure task accuracy stability. For each participant’s fMRI signal, a Dice coefficient (DC) of similarity was calculated, describing the overlap of thresholded signal (p=0.01 uncorrected) between sessions one and two. DC ranged from 0 (no overlap) to 1 (perfect overlap). DCs were calculated for voxel-wise whole-brain data and for bilateral regions of interest (frontal, temporo-parietal and motor-articulation large ROIs). Item-level naming stability was fair (M=0.34, SD=0.19) with wide range across participants (Cohen's Kappa: -0.07-0.64). Accuracy across sessions was stable (r=0.94, p<0.001). Fair reliability was shown for the group for whole-brain and ROI data (DC range: 0.2-0.39), suggesting that a picture naming task in stroke elicited poorer reliability compared to a picture naming task in neurotypical adults (i.e. Wilson et al 2016). Further, there was a wide range of participant-specific reliability in whole-brain and ROI data, in that some participants with stroke exhibited poor reliability while others exhibited good or excellent reliability. In conclusion, it is important to appreciate various aspects of behavior stability such as accuracy and item-level response, as variability may be shown in one measure and not the other. Further, fMRI reliability was highly participant-specific. Together, these results underscore the importance of multiple pre- and post-therapy assessments so that participant-specific variance is factored into outcome.

Topic Area: Methods

Back to Poster Schedule