Data Mining Approaches for Alzheimer's Disease Diagnosis

摘要

Alzheimer's disease (AD) is known for its diagnosis difficulty, we can say that if someone is suffering from Alzheimer, he could have been affected years before the diagnosis. Geriatricians are mostly confronted with a large number of patients to treat without being able to reduce their number or classify them automatically. Related to the Moroccan context, and due to magnetic resonance imaging (MRI) scan costs and MRI scanners absence in most of Moroccan regions, we choose to use clinical data to understand the disease and help classifying its subjects to increase the quality of Alzheimer's diagnosis in Morocco. This work is about the treatment of Alzheimer's clinical data, using Data Mining. We propose a model composed by three classification and prediction algorithms which are 'Decision trees', 'Discriminant analysis' and 'Logistic regression'. Our model will firstly be able to classify and categorize suffering patients (AD) from those with mild cognitive impairment (MCI) and healthy subjects (HS), secondly it will offer some affectation rules for new subjects so we can place them in the right category.