MULTICLASS DIAGNOSIS OF ALZHEIMER'S DISEASE STAGES BASED ON MULTIMODAL DATA

Alzheimer’s Disease (AD) is the leading cause of dementia worldwide, affecting memory, language, and thinking abilities. Although there is no cure, early diagnosis is essential to slow disease progression. Several studies have applied machine learning techniques for AD diagnosis; however, many rely on a single biomarker modality and perform only binary classification. This article proposes a multiclass classification approach for AD stages into three groups: AD, MCI, and HS. To achieve this, a multimodal approach combining MRI, PET, and CSF biomarkers obtained from the ADNI database was employed. Canonical Correlation Analysis (CCA) was applied for dimensionality reduction, while the SMOTETomek technique was used to address data imbalance. Subsequently, a Support Vector Machine (SVM) classifier was employed to identify the stages of Alzheimer’s Disease. The experiments resulted in the AD-M010 method, which achieved an accuracy of 92.08\% and an \textit{F1-Score} of 87.56\%, demonstrating the effectiveness of the proposed multimodal approach for multiclass classification.