Multiple kernel learning in the primal for multimodal Alzheimer's disease classification.

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Alzheimer's disease (AD) detection relies on machine learning, but faces challenges with limited training data and diverse feature representations.
• Existing methods often struggle to efficiently integrate multimodal data for accurate AD classification.

Purpose of the Study:

• To propose a novel multiple kernel-learning framework for effective and efficient Alzheimer's disease classification.
• To address the challenges of limited training samples and heterogeneous feature representations in AD detection.

Main Methods:

• A novel multiple kernel-learning framework is proposed, combining multimodal features for AD classification.
• The optimization is approached from the primal space by computing the mapping function via Fourier transform of the Gaussian kernel.
• Group lasso regularization (mixed L21 norm constraint) is applied to kernel weights for feature modality selection and exploiting complementary information.

Main Results:

• The proposed method demonstrates a straightforward solution in the primal space.
• Group sparsity enforces feature modality selection, enhancing the extraction of discriminative features.
• Experiments on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset confirm the method's effectiveness.

Conclusions:

• The novel multiple kernel-learning framework offers a scalable and implementable solution for Alzheimer's disease classification.
• The method effectively combines multimodal features and performs feature modality selection, leading to improved classification performance.
• This approach provides a promising direction for enhancing the accuracy and efficiency of Alzheimer's disease detection systems.