KERNEL-BASED MULTI-TASK JOINT SPARSE CLASSIFICATION FOR ALZHEIMER'S DISEASE

Area of Science:

• Neuroimaging
• Machine Learning
• Medical Diagnostics

Background:

• Neurodegenerative disorders like Alzheimer's disease (AD) and mild cognitive impairment (MCI) require accurate diagnostic tools.
• Multi-modality imaging, such as MRI and PET, offers complementary data for enhanced diagnostic capabilities.

Purpose of the Study:

• To develop and evaluate a novel kernel-based multi-task sparse representation model for improved classification of AD and MCI.
• To leverage the combined strengths of MRI and PET imaging features for more precise diagnosis.

Main Methods:

• Proposed a kernel-based multi-task sparse representation model integrating MRI and PET data.
• Employed multi-task learning to enforce class-level joint sparsity across imaging modalities.
• Extended the framework to the reproducing kernel Hilbert space (RKHS) to capture nonlinear feature relationships.

Main Results:

• Achieved 93.3% accuracy in classifying Alzheimer's disease (AD) from healthy controls.
• Attained 78.9% accuracy in classifying mild cognitive impairment (MCI) from healthy controls.
• Demonstrated the model's effectiveness using the Alzheimer's Disease Neuroimaging Initiative (ADNI) database.

Conclusions:

• The proposed kernel-based multi-task sparse representation model shows significant promise for the accurate classification of AD and MCI.
• Combining multi-modality imaging data through advanced sparse representation techniques enhances diagnostic performance.
• This approach offers a valuable tool for early detection and study of neurodegenerative disorders.