Multiscale Brain Structural Network Reorganization in Diabetic Retinopathy Using Regional Radiomics Similarity Networks.
Authors
Affiliations (5)
Affiliations (5)
- The Affiliated Eye Hospital, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, People's Republic of China.
- Jiangxi Province Key Laboratory of Ophthalmology and Vision Sciences, Nanchang, Jiangxi, People's Republic of China.
- Jiangxi Clinical Research Center for Ophthalmic Disease, Nanchang, Jiangxi, People's Republic of China.
- Jiangxi Provincial Key Laboratory of Vitreoretinal Diseases for Health, Nanchang, Jiangxi, People's Republic of China.
- School of Ophthalmology and Optometry, Jiangxi Medical College, Nanchang University, Nanchang, Jiangxi, People's Republic of China.
Abstract
Diabetic retinopathy (DR), a common microvascular complication of diabetes, is increasingly viewed as a systemic disorder involving the retina-brain axis. Although neuroimaging studies have reported brain abnormalities in DR, large-scale network reorganization and its biological substrates remain unclear. This cross-sectional study included 47 patients with type 2 diabetes mellitus and non-proliferative diabetic retinopathy (T2DM-NPDR) and 45 matched healthy controls. Structural MRI data were used to construct individualized regional radiomics similarity networks (R2SNs). Group differences in network architecture were examined using multiscale network analysis and gradient mapping. Epicenter mapping identified candidate regions related to network abnormalities. Imaging-transcriptomic analysis based on the Allen Human Brain Atlas, cell-type enrichment analysis, and spatial associations with normative neurotransmitter receptor and transporter maps were performed. Machine learning classifiers and an interpretable graph neural network assessed the discriminative value of R2SN-derived features. Patients with T2DM-NPDR showed widespread cortical and subcortical alterations in R2SN organization. Gradient analysis revealed a systematic shift in macroscale hierarchy, indicating disrupted cortical hierarchical organization. Epicenter mapping highlighted frontal regions, especially the inferior frontal gyrus, middle frontal gyrus, and orbitofrontal cortex, with additional involvement of the insula and parahippocampal gyrus, suggesting abnormalities beyond primary visual pathways and implicating higher-order integrative systems. Imaging-transcriptomic analysis showed that the spatial pattern of R2SN alterations was associated with gene expression profiles enriched in neuronal and glial cell populations. R2SN abnormalities also showed significant spatial correspondence with normative neurotransmitter receptor and transporter maps. Predictive models demonstrated that R2SN-derived features could distinguish patients with T2DM-NPDR from healthy controls. T2DM-NPDR is associated with multiscale structural brain network alterations beyond the visual system. Integrating radiomics-based network modeling, gradient analysis, molecular mapping, and graph-based prediction provides a framework for characterizing brain network disruption and potential biological associations in T2DM-NPDR.