Predicting Alzheimer's Disease Cognitive Assessment via Robust Low-Rank Structured Sparse Model

Area of Science:

• Neuroscience
• Medical Imaging
• Biostatistics

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder.
• Early identification of AD is critical for timely intervention.
• Existing models often overlook the complex interactions between cognitive scores and neuroimaging data.

Purpose of the Study:

• To develop a novel method for identifying informative longitudinal neuroimaging markers for early Alzheimer's disease (AD) detection.
• To predict cognitive measures by effectively utilizing neuroimaging data.
• To address the limitation of existing regression models that do not fully account for interactions between cognitive scores.

Main Methods:

• Proposed a robust low-rank structured sparse regression (RLSR) method.
• Utilized novel mixed structured sparsity-inducing norms and low-rank approximation.
• Developed an efficient algorithm with proven convergence for solving the non-smooth objective function.

Main Results:

• The RLSR method simultaneously selects effective neuroimaging features.
• The model learns the underlying structure between cognitive scores.
• Empirical studies on the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort demonstrated superior performance compared to existing methods.

Conclusions:

• The proposed RLSR method is effective for identifying neuroimaging markers and predicting cognitive decline in Alzheimer's disease.
• This approach enhances early AD detection by considering cognitive score interactions.
• The findings suggest a promising direction for developing advanced predictive models in neurodegenerative disease research.