Integrating Whole-Brain Tumor Burden MRI Features with Inflammatory Mediators to Predict High Risk of Neurocognitive Decline in Glioma Patients

Motivation: Patients with low-grade glioma (LGG) face postoperative neurocognitive decline (ND), affecting quality of life despite prolonged survival. Goal(s): The present study aims to construct and evaluate predictive models for ND in adult LGG patients. Approach: his study employed a 3D U-Net deep learning model for LGG segmentation, extracted radiomic features, and combined them with whole brain cortical features , pathological and inflammatory markers using automated machine learning to predict ND. Results: The integrated model including achieved highest predictive accuracy (AUC: 0.94), demonstrating that combining whole-brain tumor burden MRI Features, pathological, and inflammatory markers significantly improves ND prediction, enabling better risk stratification and tailored interventions. Impact: This study underscores the critical role of whole-brain tumor burden MRI features, pathological and inflammatory markers to predict ND. It encourages new research into inflammation-cognition links, promotes personalized care in glioma treatment, and could revolutionize approaches in neuro-oncology patient management.