Twin SVM-Based Classification of Alzheimer's Disease Using Complex Dual-Tree Wavelet Principal Coefficients and LDA

Area of Science:

• Neuroscience
• Medical Imaging
• Machine Learning

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder and a leading cause of dementia, significantly impacting health and socioeconomic factors.
• Early classification of AD and mild cognitive impairment (MCI) from healthy controls (HC) is crucial for implementing preventive strategies and mitigating risk factors.
• Magnetic resonance imaging (MRI) serves as a noninvasive biomarker, revealing morphometric differences and structural brain changes associated with AD.

Purpose of the Study:

• To propose and evaluate a novel approach for the accurate classification of Alzheimer's disease (AD) from healthy controls (HC).
• To leverage advanced signal processing and machine learning techniques for enhanced early detection of AD using MRI data.
• To compare the efficacy of the proposed method against existing conventional prediction techniques for AD.

Main Methods:

• Utilized dual-tree complex wavelet transforms (DTCWT) to extract principal coefficients from transaxial slices of MRI images.
• Employed linear discriminant analysis (LDA) and twin support vector machine (TWSVM) for classification of AD patients and healthy controls.
• Validated the proposed method on two independent datasets: the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the Open Access Series of Imaging Studies (OASIS).

Main Results:

• The proposed method achieved high prediction accuracy, reaching up to 92.65% ± 1.18% on the ADNI dataset and 96.68% ± 1.44% on the OASIS dataset.
• Demonstrated excellent sensitivity and specificity, with results such as 93.11% ± 1.29% sensitivity and 92.19% ± 1.56% specificity on ADNI, and 97.72% ± 2.34% sensitivity and 95.61% ± 1.67% specificity on OASIS.
• The performance metrics (accuracy, sensitivity, specificity) were found to be comparable or superior to those of various conventional AD prediction methods.

Conclusions:

• The novel approach integrating DTCWT, LDA, and TWSVM shows significant promise for accurate and early detection of Alzheimer's disease from MRI data.
• The high accuracy and robust performance across multiple datasets suggest the clinical utility of this method in distinguishing AD from HC.
• This technique offers a valuable tool for improving diagnostic capabilities in Alzheimer's disease research and potentially in clinical settings for early intervention.