Tractography-based correlation of diffusion anisotropy metrics in chronic post-stroke aphasia.

backgroundCorrelational tractography identifies white matter segments whose diffusion metrics correlate with behavioral scores, most often using fractional anisotropy (FA), while generalized FA (gFA) and quantitative anisotropy (QA) are less commonly used. PURPOSE: To investigate the relationship between anisotropy metrics and language ability, while also evaluating the ability of language-related tract metrics to classify fluent versus non-fluent post-stroke aphasia. MATERIALS AND

methodsThe FA, gFA, and QA connectometry databases were created by aggregating diffusion datasets from 33 patients. Connectometry analysis was performed to identify white matter tracts associated with repetition and naming scores. A nonparametric Spearman partial correlation was conducted, controlling for sex, age, time from onset, total intracranial volume, and lesion volume. Additionally, anisotropy metrics of language-related pathways were used to implement support vector machine and logistic regression with 5-fold stratified cross-validation for binary classification.

resultsSignificant correlations obtained from FA- and gFA-based connectome data were consistent, whereas QA-based connectome data revealed distinct correlation patterns with naming and repetition scores. Positively correlated pathways were mainly linked to left-hemisphere tracts, whereas negatively correlated pathways were largely interhemispheric connections. Our results suggest that anisotropy metrics can help distinguish fluent from non-fluent aphasia. Using the combined anisotropy metrics of language-related tracts with language scores further enhanced classification potential, especially for the right Extreme Capsule and the Corpus Callosum.

conclusionThis study highlights how different anisotropy measures yield distinct correlation patterns, demonstrating the feasibility of correlational tractography for exploring the link between white matter integrity and language ability, and suggesting that tract-specific anisotropy metrics support classification of fluent versus non-fluent aphasia.