Alzheimer’s disease (AD) is shown to be heterogeneous, displaying various subtypes with distinct characteristics. However, there is a lack of focus on the heterogeneities of the early stages of AD, which is the Mild Cognitive Impairment (MCI). Additionally, previous models had low accuracies in classifying MCI and Cognitively Normal (CN) patients. This study seeks to bridge this gap by employing a data-driven methodology, specifically clustering analysis, to explore and delineate the unique characteristics and similarities between MCI and CN groups. This study collected Magnetic Resonance Images (MRIs) of 359 MCI and 289 CN participants, their demographic information, and cognitive performance from the Alzheimer’s Disease Neuroimaging Initiative (ADNI). MRIs were preprocessed with Freesurfer to extract brain region volume features, which were then clustered by the Cancer Integration via Multikernel Learning (CIMLR) model. These features were ranked by their contributions to the clustering results. Then, statistical analysis was conducted on ranked brain features, demographics, and cognitive results to discern the clusters’ distinct patterns in brain volume characteristics. The analysis revealed five clusters, with frontal and temporal brain features, appeared mostly in the top-ranked features, while less important features were located in the optic chiasm, blood vessels, and corpus callosum regions. The clusters varied in their degree of atrophy, with Cluster 2 and 5 showing the mildest atrophy and highest percentage of CN individuals in most brain features, and Cluster 3 characterized by the most severe atrophy. The hippocampus and diencephalon region, including the thalamus and caudate, also displayed notable differences across clusters.