Related Experiment Videos
Characterizing Alzheimer's Disease With Image and Genetic Biomarkers Using Supervised Topic Models
IEEE Journal of Biomedical and Health Informatics
|August 6, 2019
Summary
This study introduces a novel supervised topic model for Alzheimer's disease (AD) characterization. The method integrates neuroimaging and genetic data to identify disease patterns and associations, aiding in diagnosis.
Area of Science:
- Neuroscience
- Medical Imaging
- Genetics
Background:
- Neuroimaging and genetic biomarkers are crucial for Alzheimer's disease (AD) diagnosis.
- Generative models offer potential for understanding AD but are underexplored.
- Supervised components can enhance generative models for discriminative characterization.
Purpose of the Study:
- To propose an original supervised topic modeling method for AD characterization.
- To leverage generative approaches for disease understanding while maintaining discriminative power.
- To jointly analyze neuroimaging and genetic data for AD classification.
Main Methods:
- Developed a supervised topic model integrating discretized image features and categorical genetic features.
- Utilized diagnostic information (cognitively normal, mild cognitive impairment, AD) as a supervision variable.
- Applied the model to the Alzheimer's Disease Neuroimaging Initiative (ADNI) cohort.
Main Results:
- The model achieved competitive discriminative performance in differentiating AD populations.
- Discovered topics revealing both known and novel neuroanatomical patterns (temporal, parietal, frontal regions).
- Identified associations between genetic factors and observed neuroanatomical patterns.
Conclusions:
- Supervised topic modeling provides a powerful generative yet discriminative approach for AD characterization.
- The method enhances understanding of AD by uncovering complex relationships between genetics and neuroanatomy.
- This approach holds promise for advancing AD diagnosis and research.