Feasibility of Alzheimer's Disease Early Detection Through Machine Learning Applied to Microwave Sensing Data Collected from a Realistic Phantom

This work aims to assess the feasibility of microwave sensing as a tool to allow early non-invasive diagnosis of Alzheimer's disease. In particular, it addresses the use of optimized support vector machine trained to perform a binary classification between the healthy and the pathological cases applied to measured scattering parameters data, collected on a realistic multi-tissue head phantom via a multiport vector network analyzer. The phantom mimics a human head in the healthy and the Alzheimer's disease condition at four different stages. The obtained experimental results are promising and show high accuracy, suggesting the feasibility of the proposed diagnostic technology.