Regional brain aging: premature aging of the domain general system predicts aphasia severity

Older age is associated with cognitive decline, higher stroke incidence, and poorer post-stroke recovery, including more severe aphasia. However, there are large interindividual differences in age-related cognitive decline and stroke recovery, often attributed to cognitive reserve. Structural brain integrity, including rates of gray matter atrophy, is a potential marker of reserve and relates to cognitive performance and recovery after stroke. Brain aging is better captured by multivariate, region-weighted measures rather than global volume alone, as some regions (e.g., prefrontal cortex) show greater age-related atrophy and others are preferentially affected in disease. Strokes may exacerbate regional atrophy and accelerate regional brain aging through direct damage, disconnection, or reduced engagement, possibly leading to asymmetric aging within a person. The authors hypothesized that atrophy and advanced regional brain age in specific brain systems (particularly domain-general regions implicated in language recovery) would be important determinants of chronic aphasia severity, beyond effects of lesion volume, chronological age, and global gray matter volume. They estimated global and regional brain age using a large normative cohort to examine relationships between regional brain aging and aphasia severity.

Prior literature links chronological aging to declines in processing speed, memory, and broader cognition, as well as worse stroke outcomes. Cognitive reserve, reflected in preserved structural integrity and lower atrophy rates, moderates age-related decline and recovery following stroke. Regional specificity of aging has been documented, with prefrontal regions showing large age-related changes, hippocampal regions more affected in Alzheimer’s, and post-stroke atrophy observed in ipsilateral thalamus and gradual hemisphere-wide shrinkage. Brain age models, derived from multivariate neuroimaging, predict mortality and cognitive outcomes and are elevated after stroke, with higher brain age associated with more severe aphasia and reduced therapy gains. White matter hyperintensities and small vessel disease also predict worse post-stroke outcomes and may contribute to tissue integrity and aging. Collectively, these findings suggest that regional, system-specific brain aging may be critical for functional outcomes after focal lesions, motivating the current focus on domain-general systems and language outcomes in chronic aphasia.

Design: Cross-sectional analysis of chronic left-hemisphere stroke survivors with aphasia and healthy controls to estimate global and regional brain age and test associations with aphasia severity. Participants: Healthy controls (n=232; 20–80 years, proficient in English) from the ABC@UofSC Repository. Stroke participants (n=89) from the POLAR clinical trial (C-STAR, USC/MUSC), with chronic aphasia (≥12 months post-stroke), left MCA ischemic or hemorrhagic stroke, age 21–80, English as primary language for ≥20 years, and able to consent. Exclusions: severely limited spontaneous speech (WAB-R 0–1), severely impaired auditory comprehension (WAB-R 0–1), bilateral/cerebellar stroke, MRI contraindications. Multiple strokes allowed if confined to left supratentorial territory. WAB-R administered by ASHA-certified SLPs; WAB AQ and subscores (naming, spontaneous speech, repetition, comprehension) computed. Imaging acquisition: Siemens 3T (Trio/Prisma). T1: MP-RAGE, 1 mm isotropic, matrix 256×256, FA 9°, 92 slices, TR 2250 ms, TI 925 ms, TE 4.11 ms. T2: 3D SPACE, 1 mm isotropic, FOV 256×256 mm, 160 sagittal slices, variable flip angle, TR 3200 ms, TE 212 ms, GRAPPA×2. Lesion processing: Lesions manually delineated on T2 by a blinded neurologist or trained staff, co-registered to T1, transformed to native T1, smoothed (3 mm FWHM). Enantiomorphic segmentation-normalization (nii_preprocess pipeline integrating SPM12, FSL, ASLtbx, MRtrix) created a chimeric T1 with mirrored healthy tissue replacing lesion, followed by unified segmentation-normalization; transforms applied to native T1, lesion, and T2/DWI. Brain age (controls): BrainAgeR pipeline on T1 images: segment gray/white matter, DARTEL normalization, QC, vectorization, PCA (top 80% variance), pretrained Gaussian process regression (Kernlab) trained on 3377 healthy scans and validated on 611 scans (ages 18–90) to estimate global brain age. This global brain age measure is used only here. Gray matter volume (all participants): VBM via CAT12/SPM12 with default steps (template registration, tissue segmentation, bias correction). Gray matter volumes (regional and total) computed using the JHU atlas ROIs and normalized by total intracranial volume. ROI grouping: 189 JHU ROIs grouped as follows: left hemisphere (94), right hemisphere (94), domain-general regions (16: left 8, right 8), language-specific regions (18: left 9, right 9), frontal (32: L16/R16), temporal (26: L13/R13), parietal (12: L6/R6), occipital (10: L5/R5). Specific membership detailed in Supplementary Data 2. Regional brain age estimation for stroke participants: For each participant and each region set (e.g., left domain-general), only spared (intact) ROIs were included. In within-range controls (controls whose global estimated brain age was within 5% of chronological age; n=126), a multiple linear regression was fit with dependent variable = global estimated brain age (BrainAgeR) and independent variables = gray matter volumes of the spared ROIs corresponding to the participant’s intact set. The participant’s regional brain age was then estimated by entering their gray matter volumes for those spared ROIs into that model. This yielded a per-participant estimate for left/right hemisphere, domain-general, language-specific, and lobar regions. The brain age gap (BrainGAP) was computed as BrainGAP = regional brain age − chronological age; positive values indicate premature brain aging. Regional brain age for controls: Out-of-range controls (n=106; global brain age >5% above/below chronological age) had regional brain age estimated via the same approach (including all ROIs, as they had no lesions) to enable group comparisons. Within-range controls were not used for regional estimation because their data formed the normative models. Statistical analyses: Conducted in MATLAB R2017b. Mixed-effects ANCOVA: 2 (group: out-of-range controls vs stroke aphasia) × 2 (hemisphere: left vs right) with dependent variable = BrainGAP in domain-general ROIs; covariates = age, education, sex, handedness. Independent-samples t-tests: compare ages of controls and stroke participants. Correlations: Pearson correlations between left and right hemisphere regional brain ages across regions; correlations between regional brain age and regional gray matter volume (z-scored; ICV-adjusted) across regions; multiple comparison correction applied (p thresholds 0.007 for 7 region pairs; 0.0036 for 14 tests). Model accuracy vs number of ROIs: correlations between number of spared ROIs used in the model and both the mean and SD of residuals (assessing accuracy). Multiple linear regressions (behavioral outcomes): Dependent variables = WAB AQ and subscores (naming, spontaneous speech, repetition, comprehension); predictors = BrainGAP, average gray matter volume (atrophy), lesion volume, participant age, and number of ROIs used; alpha adjusted for multiple comparisons (0.01).

- Mixed-effects ANCOVA (domain-general regions): No main effect of group (out-of-range controls vs stroke aphasia), F(1,263)=0.039, p=0.844. Age and education were significant covariates (age: F(1,263)=45.300, p<0.001; education: F(1,263)=12.259, p<0.001). Significant hemisphere × group interaction: F(1,263)=18.178, p<0.001. Stroke aphasia participants showed increased regional brain age (higher BrainGAP) in left hemisphere domain-general ROIs compared with out-of-range controls. - Independent t-test: Stroke aphasia participants were older than controls (M=60.60, SD=11.27 vs M=46.80, SD=17.08), t(193)=-6.517, p<0.001; ranges were comparable (controls 20–79; stroke 29–80). - Left–right hemisphere concordance: Strong positive correlations between left and right hemisphere regional brain ages across regions in aphasia participants (r=0.61 to 0.87; all p<adjusted thresholds). - Brain age vs gray matter volume: Robust negative correlations between regional brain age and regional gray matter volume across all tested regions and both hemispheres (r from −0.80 to −0.98; all p<0.0036). This indicates higher estimated brain age is associated with lower gray matter volume, though the strength varies by region. - Number of ROIs and model accuracy: Left domain-general region—negative correlation between number of spared ROIs used and SD of residuals (R=−0.79, p<0.001), but no significant correlation with mean residuals (R=−0.17, p=0.1114). Left whole hemisphere—negative correlation with SD of residuals (R=−0.94, p<0.001), no significant correlation with mean residuals (R=−0.11, p=0.308). More ROIs improved precision without systematic bias. - Multiple linear regressions (left domain-general regions): Predicting WAB AQ—BrainGAP (p=0.008), gray matter volume (p=0.009), lesion volume (p<0.001), and age (p<0.001) were significant predictors. Predicting WAB comprehension—BrainGAP (p=0.001), gray matter volume (p=0.001), and lesion volume (p<0.001) were significant predictors. Number of ROIs was significant only for spontaneous speech (p=0.007). Overall, premature regional brain aging in left domain-general regions independently relates to worse aphasia severity and comprehension, above and beyond atrophy, lesion volume, and age. - Overall pattern: In left-hemisphere stroke aphasia, intact regions of the lesioned hemisphere exhibit premature aging relative to right-hemisphere homologues and controls, indicating hemispheric asymmetry in post-stroke brain aging.

Findings support the hypothesis that regional, system-specific brain aging relates to chronic aphasia severity beyond global factors. Although stroke participants did not show a global increase in brain age relative to controls, they showed asymmetric premature aging confined to the lesioned (left) hemisphere in domain-general regions, consistent with perilesional or system-level degeneration. BrainGAP and gray matter volume independently predicted WAB outcomes, suggesting that multivariate brain age captures aspects of structural aging not fully reflected by volume alone. Lesion volume and age further contributed to aphasia severity, aligning with established predictors of post-stroke outcomes. The strong left–right concordance in regional brain age and the negative association with gray matter volume validate the regional brain age estimation, while analyses of residuals indicate more spared ROIs yield more precise estimates without bias. Results underscore the behavioral relevance of isolated aging in domain-general networks that may support language recovery when language-specific cortex is damaged. The work refines the brain age concept by demonstrating regional heterogeneity in post-stroke aging and highlights the importance of targeting specific systems in prognostication and potentially in rehabilitation strategies.

The study introduces a regional brain age framework tailored to individual lesion profiles and shows that premature aging of left-hemisphere domain-general regions predicts chronic aphasia severity and comprehension, independently of gray matter atrophy, lesion volume, and age. Intact regions within the lesioned hemisphere exhibit increased brain age, indicating that regional, isolated aging matters for language outcomes after stroke. Future research should: (1) assess language-specific regions in cohorts with different lesion distributions (e.g., right hemisphere lesions) to test their contribution; (2) incorporate markers of small vessel disease (e.g., white matter hyperintensities) and cardiovascular risk to disentangle their effects on regional aging; (3) establish thresholds for clinically meaningful BrainGAP; and (4) use longitudinal designs to quantify pre- and post-stroke aging trajectories and their relation to recovery and therapy response.

- Control participants were, on average, younger than stroke participants, although age ranges overlapped. - Regional brain age comparisons used only out-of-range controls (within-range controls formed the normative models), potentially affecting group contrasts. - Participants had large left-hemisphere lesions encompassing most language-specific regions, limiting conclusions about lesioned language regions and focusing analyses on domain-general regions; contributions of language-specific aging to behavior remain uncertain. - Single timepoint design prevents inference about pre-stroke brain age and rates of post-stroke aging. - The number of spared ROIs varied across participants; fewer ROIs reduced estimation precision (larger residual SD), and number of ROIs significantly predicted spontaneous speech. - Any positive BrainGAP was labeled as premature aging without establishing a threshold; small deviations may reflect normal variability. - Potential confounds such as small vessel disease burden were not directly modeled and could influence regional tissue integrity and aging.