Poster C63, Friday, August 17, 10:30 am – 12:15 pm, Room 2000AB
Analysis of functional connectivity furthers understanding of spontaneous speech and auditory comprehension in chronic stroke.
Helga Thors1, Brielle C. Stark1, Grigori Yourganov1, Alexandra Basilakos1, Julius Fridriksson1;1University of South Carolina
Lesion-behavior analyses in post-stroke aphasia have been core for identifying areas crucial for language processing. But, language likewise relies on areas functionally communicating as a network, and this too is interrupted post-stroke. Therefore, resting state fMRI (rsfMRI) may provide complementary and supplementary information about functional connections required for language processing. Here, we compare rsfMRI and lesion analyses in the prediction of language impairments in chronic aphasia due to stroke. METHODS: We retrospectively analyzed 92 participants with pre-morbid right-handedness and a stroke to the left hemisphere (29 F, mean age=60.47±9.36 yrs; time since stroke, M=38.85±46.53 mos; lesion volume, M=107.86±95.28cm3). Aphasia was identified using the Western Aphasia Battery-Revised (WAB-R; Kertesz, 2007). Eighteen participants had anomic aphasia; Broca's: n=32; conduction: n=9; global: n=7; Wernicke's: n=6; and 20 had no aphasia. The average aphasia severity for the whole group of participants, represented by the WAB Aphasia Quotient (AQ), was 65.15±28.64. We acquired structural (T1, T2) and rsfMRI scans, lesions were demarcated manually. We assessed brain damage and reduced functional connectivity associating with impairments on two subtests of the WAB: auditory word comprehension and spontaneous speech. To do this, we predicted normalized (z-scored) behavioral scores using multivariate support vector regression within a leave-one-participant-out framework from (a) binary lesion maps (considering only the voxels damaged in at least 10% of participants), and (b) resting-state functional connectomes. Functional connectomes were computed using the AICHA parcellation with 384 regions of interest (Joliot et al., 2015) and subsequent analyses were restricted to only cortical regions. For all analyses, we regressed out overall lesion volume; did not correct for multiple comparisons; and report the two-tailed value of prediction accuracy (the correlation between actual and predicted behavioral scores). We subsequently compared prediction accuracy using Fisher’s r-to-z transform. RESULTS: Prior to z-scoring, average score for the group on spontaneous speech (max=20) was 12.5±6.16 and auditory word comprehension (max=60), 50.36±13.83. Spontaneous speech (r=0.39, p<.0001) and auditory word comprehension (r=0.34, p=.0005) impairments were significantly predicted by reduced functional connectivity. Impaired spontaneous speech was predicted by reduced ipsilesional connectivity between supramarginal and precentral gyrus and by interhemispheric connections between frontal, frontoparietal and temporoparietal connections. Impaired auditory word comprehension was predicted by reduced ipsi- and contralesional connectivity between temporoparietal cortex. Lesion damage in insula, posterior superior temporal gyrus and inferior frontal gyrus was significantly predictive of spontaneous speech impairment (r=0.34, p<.0001) and damage to temporal (anterior, posterior) and angular gyrus predicted auditory word comprehension impairment (r=0.44, p<.0001). Spontaneous speech impairment was significantly better predicted by reduced functional connectivity (z=4.46, p<.0001), but lesion damage was a better predictor of impaired auditory word comprehension (z=24.76, p<.0001). DISCUSSION: Spontaneous speech likely relies on ipsilesional and interhemispheric connectivity. In line with the dual stream model, impaired auditory word comprehension was predicted by decreased interhemispheric connectivity but in our study more accurately predicted by regional damage. These results suggest a role for left hemisphere temporoparietal cortex in language comprehension and suggests that spontaneous speech relies on dorsal stream ipsilesional connectivity but also on the recruitment of the bilateral dorsal stream areas.
Topic Area: Methods