Predictive Model for Early Detection of Mild Cognitive Impairment and Alzheimer’s Disease

The number of people affected by Alzheimer’s disease is growing at a rapid rate, and the consequent increase in costs will have significant impacts on the world’s economies and health care systems. Therefore, there is an urgent need to identify mechanisms that can provide early detection of the disease to allow for timely intervention. Neuropsychological tests are inexpensive, non-invasive, and can be incorporated within an annual physical examination. Thus they can serve as a baseline for early cognitive impairment or Alzheimer’s disease risk prediction. In this paper, we describe a PSO-DAMIP machine-learning framework for early detection of mild cognitive impairment and Alzheimer’s disease. Using two trials of patients with Alzheimer’s disease (AD), mild cognitive impairment (MCI), and control groups, we show that one can successfully develop a classification rule based on data from neuropsychological tests to predict AD, MCI, and normal subjects where the blind prediction accuracy is over 90%. Further, our study strongly suggests that raw data of neuropsychological tests have higher potential to predict subjects from AD, MCI, and control groups than pre-processed subtotal score-like features. The classification approach and the results discussed herein offer the potential for development of a clinical decision making tool. Further study must be conducted to validate its clinical significance and its predictive accuracy among various demographic groups and across multiple sites.