Hybrid Radiomic Feature-Based Machine Learning Framework for Early Cognitive Risk Stratification Using Structural MRI

Mild Cognitive Impairment (MCI) and Alzheimer’s disease represent a growing global health burden, with early detection remaining a major clinical challenge. Structural brain alterations often precede overt clinical symptoms, highlighting the need for predictive models capable of identifying at-risk individuals during asymptomatic stages. Radiomic analysis of structural MRI offers a promising approach for quantifying subtle morphological and textural changes associated with early cognitive decline. Methods: A structured machine learning framework was developed using structural MRI data from healthy and cognitively impaired individuals. Following standardized preprocessing, multidomain radiomic features were extracted, including gray-level co-occurrence matrix (GLCM) texture features, morphological descriptors (shape and area), intensity statistics, and histogram-based metrics. Multiple supervised classifiers—decision trees, linear discriminant analysis, naïve Bayes, support vector machines (SVM), k-nearest neighbors (KNN), and neural networks—were evaluated. Model performance was assessed using 5-fold cross- validation to ensure robust generalization. Results: Among the evaluated classifiers, the neural network achieved the highest classification accuracy of 73.7%, followed closely by KNN at 72.48%. The integration of texture and morphological features improved predictive performance compared to single-domain feature approaches reported in prior studies. Results demonstrate that hybrid radiomic features enhance risk stratification capability in asymptomatic individuals. Conclusion: This study presents a radiomics-driven machine learning framework for early cognitive risk prediction using structural MRI. The findings support the feasibility of non-invasive MRI- based risk stratification and highlight the importance of multidomain feature integration in improving classification performance. The proposed framework may contribute to preventive screening strategies and early clinical intervention planning.