Comparing different algorithms for the course of Alzheimer's disease using machine learning

Area of Science:

• Neuroimaging
• Machine Learning
• Neurology

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder impacting memory, thinking, and language.
• Brain Magnetic Resonance Imaging (MRI) offers characteristic indexes valuable for AD assessment.
• Early and accurate diagnosis of AD and its precursor stages, like Mild Cognitive Impairment (MCI), is crucial for effective management.

Purpose of the Study:

• To evaluate the efficacy of machine learning algorithms in classifying and predicting the progression of Alzheimer's disease using MRI data.
• To identify the optimal machine learning model for auxiliary diagnosis of AD based on characteristic MRI indexes.
• To compare the performance of Random Forest (RF), Decision Tree (DT), and Support Vector Machine (SVM) algorithms in differentiating AD stages.

Main Methods:

• Utilized data from 560 subjects in the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.
• Classified subjects into four groups: Cognitive Normal (CN), Early Mild Cognitive Impairment (EMCI), Late Mild Cognitive Impairment (LMCI), and Alzheimer's Disease (AD).
• Applied RF, DT, and SVM algorithms to MRI characteristic indexes for classification and prediction, comparing accuracy, sensitivity, specificity, and AUC.

Main Results:

• The Random Forest (RF) classifier achieved the highest classification accuracy (73.8%) among the evaluated machine learning algorithms.
• RF demonstrated superior performance in predicting transitions between disease stages, particularly between Cognitive Normal (CN) and Alzheimer's Disease (AD) (AUC=0.92).
• Classification accuracy varied across disease stages, with CN-AD showing the highest accuracy, followed by EMCI-AD and CN-LMCI.

Conclusions:

• The Random Forest (RF) classifier is highly effective for classifying different stages of Alzheimer's disease, including early-stage detection.
• Specific MRI indexes, when utilized as features in the RF model, yield optimal prediction outcomes for AD progression.
• The RF classifier shows significant potential as an auxiliary tool to assist clinicians in the early diagnosis and classification of AD.