Exploring the Links Between Lexical Production and Track-Weighted Imaging in Middle-Aged Adults

1595 

Abstract Submission 

Clément Guichet¹, Arnaud Attyé², Elise Roger^3,4, Sophie Achard⁵, Martial Mermillod¹, Monica Baciu¹

¹Univ. Grenoble Alpes, LPNC, CNRS UMR 5105, Grenoble, France, ²GeodAIsics, Grenoble, France, ³Institut Universitaire de Gériatrie de Montréal, Communication and Aging Lab, Montreal, Quebec, Canada, ⁴Faculty of Medicine, University of Montreal, Montreal, Quebec, Canada, ⁵Univ. Grenoble Alpes, LJK, UMR CNRS 5224, Grenoble, France

Clément Guichet  
Univ. Grenoble Alpes, LPNC, CNRS UMR 5105
Grenoble, France

Arnaud Attyé  
GeodAIsics
Grenoble, France

Elise Roger  
Institut Universitaire de Gériatrie de Montréal, Communication and Aging Lab|Faculty of Medicine, University of Montreal
Montreal, Quebec, Canada|Montreal, Quebec, Canada

Sophie Achard  
Univ. Grenoble Alpes, LJK, UMR CNRS 5224
Grenoble, France

Martial Mermillod  
Univ. Grenoble Alpes, LPNC, CNRS UMR 5105
Grenoble, France

Monica Baciu  
Univ. Grenoble Alpes, LPNC, CNRS UMR 5105
Grenoble, France

Cognitive aging manifests as a gradual deterioration in lexical retrieval and generation, referred to as lexical production (LP), beginning in middle age (1). Previous research showed that LP decline is not solely attributed to domain-specific mechanisms (e.g., language), but also involves domain-general factors like executive functioning (2). While disruptions in functional connectivity have been linked to age-related LP decline (3), the structural correlates remain unclear. Therefore, this study aims to investigate structural modifications in conjunction with cognitive scores to elucidate LP decline among middle-aged adults.

We analyzed 7 cognitive scores and diffusion MRI data obtained from 155 healthy adults aged 45 to 60 in the CAMCAN cohort (4). T1w MR images were preprocessed using FreeSurfer's recon-all command. The standard mrtrix3 pipeline was applied for diffusion MR preprocessing and whole-brain tractography (SIFT2; 5). We quantified structural modifications with Track-Weighted Imaging (TWI; 6). To establish voxel-wise correspondence across subjects, each subject's tractogram was first registered to a study-based population template and a TW-FA map was built with the subject's fractional anisotropy (FA) image previously warped to template space. Structure-cognition links were modeled using Partial Least Square Correlation (PLSC) contrasted across three age groups (45-50; 51-55; 56-60). This contrast ensured that covariance effects were isolated from absolute age-related differences in either modality (7). To start the PLSC, TW-FA maps were vectorized and stacked across subjects. This TW-FA matrix was cross-correlated with the matrix containing cognitive scores, yielding a covariance matrix that was submitted to singular value decomposition (Figure 1).

One latent component, capturing 32% of the covariance between TW-FA values and cognitive scores (singular value = 6,845.4, p = .01), distinguished individuals aged 56 and above from the other two groups (BSR56-60>45-55 = 20.7, BSR51-55>45-50 = 0.9; Figure 2A). This component was associated with lower cognitive performances (tip-of-the-tongue = -8.3; naming = -8.9; sentence comprehension = -7.8, proverb = -2.7, Cattell = -8.4), and lower TW-FA values predominantly in clusters located in the right inferior fronto-temporal white matter (WM; Figure 2B). Surprisingly, increased TW-FA in the WM adjacent to the right temporo-parietal junction (TPJ) also contributed to the cognitive decline pattern described above. Exploratory analysis showed that salient voxels (BSR >= 2.58) in the right TPJ are crossed by bundles with different orientations: association (52.2% of the right middle longitudinal fascicle, MLF; 27.3% of the right arcuate fasciculus: AF), callosal (34.8% of the isthmus, CC6), and to a lesser extent by fibers originating from the striatum (21.8%), thalamus (20.2%), and occipital lobe (16.9%) (Figure 2C).

Our study reveals a pattern of covariance between brain structure and lexical production which distinguishes middle-aged adults before and after 56 yo. Although reduced WM integrity around fronto-temporal regions is mainly predictive of lower LP performances, the same alterations also contribute to the age-related decline in domain-general processes. Interestingly, the degree of TW-FA in the right TPJ could be a potential structural biomarker for LP performance in middle-aged adults. Notably, a recent work argues that increased TW-FA during aging, as reported in our study, is not related to better WM integrity but to selective fiber degeneration (8). This hypothesis is supported by our study's identification of the bundles converging to the right TPJ, which have been reported to form a bottleneck region of crossing fibers along an anterior-posterior axis (9). However, we acknowledge that our voxel-based analysis is insufficient for distinguishing fiber differences in WM alterations in this bottleneck region. Further research to validate these findings is planned.

Speech Production

Aging ²

Diffusion MRI Modeling and Analysis ¹

Multivariate Approaches

White Matter Anatomy, Fiber Pathways and Connectivity

Aging

Cognition

Language

Multivariate

Tractography

WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC