Subregional Limbic Radiomics on FDG-PET Provides Accurate Early Detection of Alzheimer's Disease

Background To investigate the radiomics features of the hippocampus and the amygdala subregions in FDG-PET images that can best differentiate Mild Cognitive Impairment (MCI), Alzheimer's Disease (AD), and healthy patients. Methods Baseline FDG-PET data from 555 participants in the ADNI dataset were analyzed, comprising 189 cognitively normal (CN) individuals, 201 with MCI, and 165 with AD. We extracted 120 quantitative features from finely and automatically parcellated subregions (hippocampal n  = 38, amygdala n  = 18) using a probabilistic atlas. To identify the most effective classification model, we applied four feature selection techniques, ANOVA, PCA, LASSO, and Chi-square, combined with nine different classifiers, resulting in 36 unique model combinations. This comprehensive evaluation enabled the selection of a high-performing machine learning pipeline. Results The Multi-Layer Perceptron (MLP) model combined with LASSO demonstrated excellent classification performance: ROC AUC of 0.957 for CN vs. AD, ROC AUC of 0.867 for MCI vs. AD, and ROC AUC of 0.782 for CN vs. MCI. Key regions, including the accessory basal nucleus, presubiculum head, and CA4 head, were identified as critical biomarkers. Features including GLRLM (Long Run Emphasis) and Small Dependence Emphasis (GLDM) showed strong diagnostic potential, reflecting subtle metabolic and microstructural changes often preceding anatomical alterations. Conclusions Specific hippocampal and amygdala subregions and their four radiomic features were found to have a significant role in the early diagnosis of AD, its staging, and its severity assessment by capturing subtle shifts in metabolic patterns. Furthermore, these features offer potential insights into the disease's underlying mechanisms and model interpretability.