This study leverages Diffusion Tensor Imaging (DTI) data from the ADNI database to classify Alzheimer’s Disease (AD) stages—Early Mild Cognitive Impairment (EMCI), Late Mild Cognitive Impairment (LMCI), and AD—while distinguishing them from healthy controls (HC). A total of 228 features were extracted from 57 regions of interest (ROIs) and analyzed using advanced machine learning methods. Classification models were rigorously trained and validated through 10-fold cross-validation. The best-performing model achieved a test accuracy of 94.3% in distinguishing the four groups (HC, EMCI, LMCI, and AD), underscoring the importance of model selection and feature engineering. The Areas Under the Curve (AUCs) were 0.97, 0.96, 0.99, and 0.91 for HC, EMCI, LMCI, and AD, respectively. Feature ranking highlighted Axial Diffusivity (AxD) in the uncinate fasciculus as a key biomarker with strong discriminative power across all stages of AD. Moreover, regions such as the sagittal stratum and hippocampal cingulum were also implicated, reflecting their known roles in memory and executive function decline. Overall, these findings enhance understanding of the neuropathological mechanisms underlying AD and demonstrate the potential of DTI-based machine learning as a non-invasive tool for early diagnosis and personalized treatment planning.