Cognitive aging in healthy adults exhibits significant and heterogeneous variability. In this study, we apply a robust unified learning framework to cluster subgroups using neuroimaging data (brain volume and white matter), to identify neurological phenotypes that can sort out the heterogeneity in cognitive aging and help identify potential risk factors for suboptimal brain aging. Using machine learning analytics, results revealed two unique subgroups in healthy older adults with different patterns of white matter integrity and brain volumetric measures. The classification of phenotypical subgroups in healthy older adults may inform the understanding of the complexity of brain changes before the onset of clinical symptoms. The identified neuroimaging features that defined group classification are recognized as important structures that subserve cognitive performance. Further analysis of these potential biomarkers that help predict trajectory of cognitive decline in symptom free individuals could lead to the detection of early stages of neurodegenerative diseases.