Federated Learning Model for Early Detection of Dementia Using Blood Biosamples
Mohamed Elsersy, Ahmed Sherif, Ahmad Imam, Mohammad Mahbubur Rahman Khan Mamun, Kasem Khalil, Mohamed Haitham
Abstract
Alzheimer’s disease (AD) is a serious, long-term health problem that causes much pain and loss for the person with it and their family. Its early and accurate detection might result in a substantial reduction of the disease outcomes and consequences. Blood biosamples are a simple and inexpensive technique in medical testing. This paper proposes diagnostic models for blood biosamples based on federated learning (FL) and its modifications to detect AD early. Our experiments used blood biosample data sets from the ADNI website to evaluate our models. Our performance analysis indicates that our algorithms are more accurate and achieve an accuracy of 87% for early detection.