Normal brain aging and Alzheimer's disease both involve progressive structural brain alterations, making it challenging to distinguish pathological neurodegeneration from normative aging-related atrophy. This study investigated whether Alzheimer's disease exhibits radiomic patterns that mimic, diverge from, or selectively perturb age-associated structural brain changes. T1-weighted magnetic resonance imaging scans from the Alzheimer's Disease Neuroimaging Initiative were analyzed using a region-wise radiomics framework across 10 anatomically defined brain regions. Radiomic features were extracted following automated segmentation, bias field correction, and intensity normalization. Age-associated radiomic patterns were first identified in cognitively normal subjects using Spearman correlation analysis. Features demonstrating significant age sensitivity were subsequently compared between cognitively normal and Alzheimer's disease cohorts across age bins using Welch's two-sample t-tests with permutation-based significance estimation and false discovery rate correction. Medial temporal and limbic regions, particularly the hippocampus, entorhinal cortex, and cingulum, demonstrated consistent age-aligned radiomic trajectories with systematic, statistically significant disease-related shifts across all age bins, supported by large effect sizes and bootstrap-validated confidence intervals. In contrast, several other regions demonstrated more heterogeneous and less stable patterns of group separation across age bins. Secondary analysis using late mild cognitive impairment subjects demonstrated that these radiomic divergences are detectable at the transition from normal cognition to mild cognitive impairment, with statistically significant CN-LMCI separation but no significant LMCI-AD separation, positioning the identified markers as early-stage rather than late-stage indicators of neurodegeneration. These findings indicate that Alzheimer's disease does not uniformly mimic normal aging across the brain but instead selectively perturbs radiomic features associated with normative aging trajectories. The identified markers represent promising candidates for age-adjusted radiomic biomarkers, warranting validation in independent cohorts to establish their generalisability. The fully automated nature of the analytical pipeline - spanning segmentation, feature extraction, and statistical comparison without manual annotation - may facilitate scalable validation of these biomarkers in larger neuroimaging cohorts.
Sharma, M. S., Agarwal, R., Tiwari, N., Sharma, M., Kaushik, A.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 1
- Comments 0