Towards Precision Dementia Detection: Integrating ML and Clinical Data

The dominant pathology theory holds that dementia results from amyloid buildup, followed by tau protein aggregation and neurodegeneration. Current anti-amyloid and anti-tau therapies, however, show varied therapeutic success. Atypical Alzheimer's disease has radiologic and clinical symptoms with other medical conditions, making accurate diagnosis even more challenging. Our approach employs machine learning algorithms to analyze a dataset that includes clinical evaluations, neuropsychological assessment scores, and biomarker measures. The information was gathered from a broad group of people with varying cognitive states, including those in the preclinical and moderate cognitive impairment (MCI) phases as well as AD patients. This study has important implications for clinical practice, permitting prompt therapies, and advancing our understanding of the progression of Alzheimer's disease. When integrated with Boruta as a feature selector, random forest outperformed upon fine-tuning on both datasets.