You are viewing the SNL 2018 Archive Website. For the latest information, see the Current Website.

Poster B44, Thursday, August 16, 3:05 – 4:50 pm, Room 2000AB

Data-driven Aphasia Sub-Typing using Lesion-Symptom Mapping and Community Detection Analysis

Jon-Frederick Landrigan1, Daniel Mirman2;1Drexel University, 2University of Alabama at Birmingham

Models of aphasia sub-types typically focus on the distinction between the production and comprehension of speech. However, recent studies suggest that this fails to account for the full spectrum of deficits associated with aphasia and there is inconsistent relationships between lesion site and aphasia sub-type. The accumulation of data from people with aphasia in publicly available databases, combined with the development of advanced analytical techniques, provide a new opportunity to investigate the neural correlates of aphasia and language processing. The current project consisted of two parts, each of which combined two analysis methods: lesion-symptom mapping and community detection analysis (CDA), a clustering technique from network science that attempts to uncover groups of nodes in a network that are densely connected to each other and sparsely connected to other groups. The first used CDA to cluster 134 patients based on their behavioral deficit profiles across 20 different language assessments and identified three clusters: patients in cluster 1 had relatively mild deficits as compared to those in clusters 2 and 3. Patients in cluster 2 had primarily phonological deficits and patients in cluster 3 had primarily semantic deficits. Voxel-based lesion-symptom mapping (VLSM) comparisons were used to identify the lesion correlates of each cluster. Cluster 1 was associated with damage to a number of areas spanning from frontal to parietal regions. Cluster 2 was primarily associated with damage to the supramarginal gyrus extending anteriorly into the postcentral gyrus and cluster 3 was primarily associated with damage to frontal areas. The second part worked in the opposite direction. First VLSM analyses were used to identify the lesion correlates of deficits on the same 20 measures of language performance, including a general aphasia assessment (WAB), measures of speech production and comprehension, and semantic processing. These results were mapped to the Human Connectome Project template of 180 left hemisphere regions within the lesion territory. Then CDA was run to find clusters of neural regions that showed similar associations between lesion status and behavioral performance. This CDA identified 3 distinct clusters of neural regions. Cluster 1 primarily consisted of frontal regions and was associated with deficits on measures of verbal semantics (e.g., semantic errors in picture naming, synonym judgments). Regions in cluster 2 were typically found in peri-Sylvian and parietal areas and were associated with phonological processing and cluster 3 was comprised of frontal regions that were associated with deficits on visual semantics and/or semantic control (Camel and Cactus Test). Importantly, the results of these two analyses align at both the behavioral and neural level: the primary distinction was between semantic and phonological processing, with frontal regions critical for the former and peri-Sylvian regions critical for the latter. These results suggest that models of language processing and aphasia should focus first on semantic and phonological processing and then their downstream affects on speech production and comprehension.

Topic Area: Language Disorders