Developing an Artificial Intelligence Framework for Identifying Fusion Blood-Based Biomarkers in Alzheimer's Disease

Abstract

Alzheimer’s Disease (AD) is an irreversible neurological disorder, a major cause of disability among the elderly, with no effective therapeutic options currently available. It is an asympto-matic disease in the prodromal stages and begins many years before clinical appearances. Early diagnosis of AD allows patients to obtain appropriate healthcare assistance, accelerating the de-velopment of new medications. A biomarker that evaluates the alterations in the brain cells produced by AD in its preliminary periods might be significant for its early identification. Blood-based biomarkers (BBBMs) facilitate the early detection of AD. The BBBMs detection procedure is cost-efficient and minimally invasive. The aim of this study is to identify the best BBBMs, and machine learning (ML) algorithms play a significant role in identifying people at the high-risk of AD. A total of 146 BBBMs from a database by ADNI, and 12-ML algorithms were in-vestigated. The results show that linear discriminant analysis, Naive Bayes, and support vector machine are the promising ML algorithms for AD detection that integrated into the novel en-semble voting detection model. Furthermore, the four BBBMs i.e., Immunoglobulin M (IGM), Placenta Growth Factor (PLGF), Serum Glutamic Oxaloacetic Transaminase (SGOT), and Al-pha-1-Microglobulin (A1Micro) are the significant biomarkers to detect AD in its early stages with performance of 92.86% for sensitivity and 82.35% for specificity. Consequently, BBBMs are the preferred option in clinical practice. In addition, integrating artificial intelligence such as ML into healthcare might help with early detection of AD.