# Altered Functional Specialization and Interhemispheric Coordination in Rhegmatogenous Retinal Detachment: Associations With Gene Expression, Neurotransmitter Receptor Distribution, and SVM–SHAP Classification: A Multimodal Neuroimaging–Transcriptomics Study Integrating Functional Metrics and Interpretable Machine Learning

## TL;DR

This study explores how retinal detachment affects brain function, linking changes to gene activity and neurotransmitter systems using advanced imaging and machine learning.

## Contribution

The study introduces a novel multimodal approach combining neuroimaging, transcriptomics, and interpretable machine learning to explore brain changes in retinal detachment.

## Key findings

- RRD patients showed altered brain function in the frontal, occipital lobes, and thalamus.
- Genes linked to these changes are involved in synaptic signaling and neuroplasticity.
- The SVM–SHAP model identified thalamic connectivity as a key biomarker for RRD.

## Abstract

Previous studies have reported functional alterations in the brains of patients with rhegmatogenous retinal detachment (RRD). However, it remains largely unclear whether RRD affects hemispheric specialization and interhemispheric coordination, and how these alterations relate to underlying gene expression patterns and neurotransmitter receptor distributions.

We employed the Autonomy Index (AI) and Connectivity between Functionally Homotopic Voxels (CFH) to quantify alterations in hemispheric specialization and interhemispheric cooperation in patients with RRD. Transcriptome–neuroimaging spatial correlation analysis was performed by integrating gene expression data from the Allen Human Brain Atlas (AHBA) to identify genes associated with AI and CFH alterations. Enrichment and protein–protein interaction analyses were conducted to characterize the biological processes and molecular features of these genes. Furthermore, we explored the spatial associations between AI/CFH abnormalities and neurotransmitter receptor distributions. Finally, a support vector machine (SVM) classifier combined with Shapley additive explanations (SHAP) was implemented to distinguish RRD patients from healthy controls (HCs) and to determine the most discriminative brain regions.

RRD patients exhibited significant alterations in AI and CFH within the frontal lobe, occipital lobe, and thalamus. Transcriptome–neuroimaging integration revealed gene sets closely associated with these abnormalities. These genes were primarily enriched in key biological processes including synaptic signaling, sensory organ development, Notch signaling, and structural neuroplasticity. The spatial pattern of CFH changes showed strong alignment with the regional distributions of multiple neurotransmitter systems, particularly serotonergic, dopaminergic, glutamatergic, and cholinergic pathways. Finally, the SVM–SHAP classification framework identified CFH in the right thalamus as the most discriminative feature for differentiating RRD patients from HCs.

These findings deepen our neurobiological understanding of RRD‐induced brain functional remodeling and provide theoretical support and a methodological foundation for developing central intervention strategies and potential discriminative imaging tools for retinal diseases.

Autonomy Index (AI) and Connective Flexibility of Homotopic voxels (CFH) were derived from resting‐state fMRI data following standard preprocessing. These functional metrics were computed based on the AAL90 atlas to characterize intrinsic activity and interhemispheric coordination patterns in RRD patients. Using the Allen Human Brain Atlas (AHBA), regional gene expression profiles from the left hemisphere were aligned to the AAL90 template, generating a gene expression matrix encompassing 15,632 genes for transcriptomic integration. Partial least squares (PLS) regression was performed to relate spatial patterns of AI and CFH to regional gene expression. Genes contributing most significantly to the PLS components were subjected to downstream analyses including gene enrichment analysis, protein–protein interaction (PPI) network construction, and spatiotemporal‐specific expression profiling. Spatial correlations were assessed between AI/CFH alterations and neurotransmitter receptor distribution maps, offering insights into potential neurochemical mechanisms linked to functional abnormalities in RRD. Functional alterations in AI and CFH were used as input features for support vector machine (SVM) classification to distinguish RRD patients from controls. SHAP analysis was applied to interpret the contribution of each feature, enhancing model transparency and biological interpretability.