Search

Assessment of regional language lateralization is crucial in many scenarios, but not all populations are suited for its evaluation via task‐functional magnetic resonance imaging (fMRI). In this study, the utility of structural connectome features for the classification of language lateralization in the anterior temporal lobes (ATLs) was investigated. Laterality indices for semantic processing in the ATL were computed from task‐fMRI in 1038 subjects from the Human Connectome Project who were labeled as stronger rightward lateralized (RL) or stronger leftward to bilaterally lateralized (LL) in a data‐driven approach. Data of unrelated subjects (n = 432) were used for further analyses. Structural connectomes were generated from diffusion‐MRI tractography, and graph theoretical metrics (node degree, betweenness centrality) were computed. A neural network (NN) and a random forest (RF) classifier were trained on these metrics to classify subjects as RL or LL. After classification, comparisons of network measures were conducted via permutation testing. Degree‐based classifiers produced significant above‐chance predictions both during cross‐validation (NN: AUC–ROC[CI] = 0.68[0.64–0.73], accuracy[CI] = 68.34%[63–73.2%]; RF: AUC–ROC[CI] = 0.7[0.66–0.73], accuracy[CI] = 64.81%[60.9–68.5]) and testing (NN: AUC–ROC[CI] = 0.69[0.53–0.84], accuracy[CI] = 68.09[53.2–80.9]; RF: AUC–ROC[CI] = 0.68[0.53–0.84], accuracy[CI] = 68.09[55.3–80.9]). Comparison of network metrics revealed small effects of increased node degree within the right posterior middle temporal gyrus (pMTG) in subjects with RL, while degree was decreased in the right posterior cingulate cortex (PCC). Above‐chance predictions of functional language lateralization in the ATL are possible based on diffusion‐MRI connectomes alone. Increased degree within the right pMTG as a right‐sided homologue of a known semantic hub, and decreased hubness of the right PCC may form a structural basis for rightward‐lateralized semantic processing. [ABSTRACT FROM AUTHOR]