Multi-modal feature selection with anchor graph for Alzheimer's disease

Area of Science:

• Neurology
• Biomedical Informatics
• Machine Learning

Background:

• Early diagnosis of Alzheimer's disease (AD) is crucial for effective treatment and delaying disease progression.
• Current multi-modal feature selection methods for AD diagnosis often overlook inter-modal relationships and local data structures.
• Existing algorithms primarily focus on internal information within individual data modalities.

Purpose of the Study:

• To propose a novel multi-modal feature selection algorithm for the early diagnosis of Alzheimer's disease.
• To address limitations of existing methods by incorporating inter-modal relationships, modality importance, and local data structure.
• To enhance the accuracy and robustness of early AD detection using multi-modal data.

Main Methods:

• Developed a multi-modal feature selection algorithm incorporating an anchor graph approach.
• Utilized least square loss and l2,1-norm to determine feature weights within each modality.
• Integrated a modal weight factor to ascertain the importance of each data modality.
• Employed anchor graphs to efficiently capture local structure information within multi-modal data.

Main Results:

• The proposed algorithm effectively learns feature importance across multiple modalities.
• It successfully captures the relationships between different data modalities.
• The anchor graph component efficiently models local structures in the multi-modal dataset.
• Validation on the ADNI dataset demonstrated the algorithm's effectiveness.

Conclusions:

• The novel multi-modal feature selection algorithm shows promise for improving early Alzheimer's disease diagnosis.
• Considering inter-modal relationships and local structures enhances diagnostic accuracy.
• The method provides a robust framework for analyzing complex multi-modal data in neurodegenerative disease research.