Multifunctional Roles of the Ventral Stream in Language Models: Advanced Segmental Quantification in Post-Stroke Aphasic Patients

Abstract

In the dual-route language model, the dorsal pathway is known for sound-to-motor mapping, but the role of the ventral stream is controversial. With the goal of enhancing our understanding of language models, this study investigated the diffusion characteristics of candidate tracts in aphasic patients. We evaluated 14 subacute aphasic patients post-stroke and 11 healthy controls with language assessment and diffusion magnetic resonance imaging. Voxel-based lesion-symptom mapping found multiple linguistic associations for the ventral stream, while automated fiber quantification (AFQ) showed, via reduced fractional anisotropy (FA) and axial diffusivity with increased radial diffusivity (all corrected p < 0.05), that the integrity of both the left dorsal and ventral streams was compromised. The average diffusion metrics of each fascicle provided by AFQ also confirmed that voxels with significant FA-language correlations were located in the ventral tracts, including the left inferior fronto-occipital fascicle (IFOF) (comprehension: r = 0.839, p = 0.001; repetition: r = 0.845, p = 0.001; naming: r = 0.813, p = 0.002; aphasia quotient: r = 0.847, p = 0.001) and uncinate fascicle (naming: r = 0.948, p = 0.001). Furthermore, point-wise AFQ revealed that the segment of the left IFOF with the strongest correlations was its narrow stem. The temporal segment of the left inferior longitudinal fascicle was also found to correlate significantly with comprehension (r = 0.663, p = 0.03) and repetition (r = 0.742, p = 0.009). This preliminary study suggests that white matter integrity analysis of the ventral stream may have the potential to reveal aphasic severity and guide individualized rehabilitation. The left IFOF, specifically its narrow stem segment, associates with multiple aspects of language, indicating an important role in semantic processing and multimodal linguistic functions.

Keywords: aphasia; diffusion tensor imaging; language model; stroke; ventral pathway.

Figure 1

Three-level strategy for neuroimaging analysis. DTI, diffusion tensor imaging; MR, magnetic resonance; AFQ, automated fiber quantification; TBSS, tract-based spatial statistics.

Figure 2

Lesion overlay map for all the aphasic patients. Color bar indicates percentage of patients with lesions in a particular voxel (0–100%).

Figure 3

TBSS correlation analysis between the FA value and language subcomponents in aphasic patients. The green lines in every image represent a mean white matter skeleton. The regions in pink stand for AQ-related tracts with significant positive correlation (p < 0.05; TFCE corrected). Similarly, regions of significant positive correlation with repetition are in red, naming in blue, and comprehension in magenta. All the correlations were controlled by additional covariates including age, sex, and lesion size. TBSS, tract-based spatial statistics; IFOF, inferior fronto-occipital fascicle; ILF, inferior longitudinal fascicle; UF, uncinate fascicle; AIC, anterior internal capsule; ATR, anterior thalamic radiation; AQ, aphasia quotient; TFCE, threshold-free cluster enhancement; FA, fractional anisotropy.

Figure 4

Point-wise comparisons for diffusion measurement along the tract trajectory between patients and controls. The horizontal scale represents 100 equidistant nodes along the central portion of the tract, defined by the red starting point and blue point. The blue solid line refers to the mean value of patient group, while the green solid line refers the mean value of control group. The dash lines stand for their SD. The vertical scale refers to diffusion metrics, and the segments with significant difference are marked in different colors, yellow for FA, light blue for AD, and pink for RD (all corrected p < 0.05). Values of AD and RD are measured in square millimeter per second × 10⁻³. FA, fractional anisotropy; RD, radial diffusivity; AD, axial diffusivity; IFOF, inferior fronto-occipital fascicle; ILF, inferior longitudinal fascicle; UF, uncinate fascicle; SLF, superior longitudinal fascicle; AF, arcuate fascicle.

Figure 5

Correlations between different language subcomponents and mean FA value of the left major fascicles. (A) Correlation matrix with a color scale, squares in red mean positive correlations while blue ones mean negative correlations, asterisks denote significant correlations (all corrected p < 0.05). (B) Correlation scatter plots for the left IFOF, presenting the relationships between FA values (x-axis) and Z-scores of language subtests on the y-axis. FA, fractional anisotropy; IFOF, inferior fronto-occipital fascicle; ILF, inferior longitudinal fascicle; UF, uncinate fascicle; SLF, superior longitudinal fascicle; AF, arcuate fascicle; AQ, aphasia quotient.

Figure 6

Point-wise correlations along the trajectory of the left IFOF between FA value and language subcomponents. (A) Colors correspond to the magnitude of correlations along the 100 equidistant points, and red arrows points out the location of the maximal correlation; (B) the sections of significant correlations along the 100 equidistant points of the left IFOF (all corrected p < 0.05). For naming (blue), repetition (red), and reading (green), the significant sections are the narrow stem and frontal radiation; the significant sections for comprehension are within the narrow stem; (C) scatter plots with the regression curves for the point of maximal correlation, presenting the linear relationships between FA value (x-axis) and Z-scores of language subtests on the y-axis (*p < 0.05; **p < 0.01; ***p < 0.001). FA, fractional anisotropy; IFOF, inferior fronto-occipital fascicle.