Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Multivariate multichannel morphometry reveals mesoscale white matter structures supporting language processing in post-stroke aphasia
Poster Session D, Saturday, October 26, 10:30 am - 12:00 pm, Great Hall 3 and 4
Andrew DeMarco1, Peter Turkeltaub1,2; 1Georgetown University, 2MedStar National Rehabilitation Hospital
The language network is now understood to include more gray matter regions than classical models suggested, but the anatomy of white matter connections between these regions remains underspecified. Modern macroscale investigations focus on discrete, long-distance paths abstracted via tractography, with microstructure approaches like TBSS typically restricted to strictly-thresholded voxels. These studies generally ignore white matter mesoscale architecture, including gyral blades, local U-shaped fibers, and columnar cortical structures. Here, we examine the contributions of mesoscale white matter structures to language comprehension and production in a cohort of chronic left-hemisphere stroke survivors. We employ a multivariate, multichannel morphometry approach that fuses multiple DTI scalar maps to enhance interpretability of results. Behavioral scores and neuroimaging data, including diffusion imaging, from chronic stroke survivors (N=134) enrolled in prior studies in the Cognitive Recovery Lab were analyzed. Voxelwise tensors were fitted using FSL and warped to a shared template using ANTs. Three analyses employed a multivariate lesion-symptom mapping approach, with predictor features consisting of left-hemisphere voxels concatenated across the three orthogonal scalar FA maps. For proof-of-concept, we first localized visual field laterality index scores (N=96). We then localized the Auditory Verbal Comprehension (AVComp) and the Spontaneous Speech (SpontSpeech) subscores from the Western Aphasia Battery. Statistical significance was established via 5000 permutations, with multiple comparisons corrected by controlling for continuous family-wise error rate (v=1000, FWER = .05). Lesion volume was included as a covariate. Model solutions were backprojected and visualized as a fused 3D RGB-encoded interactive isosurface. Results were first interpreted via overlap with the XTRACT atlas. Critically, detailed spatial geometries were further interpreted. Patients scored an average visual field LI of -.08 (SD=.19), 84% (SD=17%) for AVComp, and 74% (SD=25%) for SpontSpeech. Significant field cut LI results (2149 voxels) showed strong convergence with the left optic radiation parcel, with a clear Meyer’s loop, spanning from calcarine sulcus around the lateral ventricle and sweeping medially toward LGN, plus vertical occipital fasciculus. SpontSpeech results (3577 voxels) mostly overlapped with superior/anterior thalamic radiations, acoustic radiations, uncinate, corticospinal tract, and medio-dorsal longitudinal fasciculus, but excluded the superior longitudinal and arcuate fasciculi. Additionally, discrete structured tendrils reached from long tracts towards middle/inferior temporal and inferior frontal gyrus cortex, with a U-shaped structure reaching into premotor cortex. AVComp results (2515 voxels) overlapped with medio-dorsal and inferior fronto-occipital fasciculi, acoustic radiations, inferior longitudinal fasciculus and superior thalamic radiation, with some uncinate and arcuate overlap. The result dipped into the anterior temporal pole, and notable geometric structures include a clear impression of the transverse temporal gyrus and approximately five discrete lateral (red) projections into the lateral temporal cortex. Here, we show proof-of-concept for mapping mesoscale white matter structures associated with language processing. This investigation found mostly expected results with regard to long-range pathways, although some surprises such as lack of dorsal pathway with spontaneous speech. The results also contain rich mesoscale white matter geometries stereotyped across individuals, which can be studied to elaborate on general dual-stream conceptualizations of white matter contributions to language processing.
Topic Areas: Disorders: Acquired, Methods