Incorporating privileged genetic information for fundus image based glaucoma detection.

Area of Science:

• Ophthalmology and Genetics
• Medical Imaging Analysis

Background:

• Retinal fundus images are commonly used for glaucoma detection due to ease of acquisition.
• Single nucleotide polymorphisms (SNPs) show promise for glaucoma detection, potentially outperforming fundus images.
• A practical challenge exists where training data includes both fundus images and genetic data, but test data only includes fundus images.

Purpose of the Study:

• To develop a method for glaucoma detection that leverages genetic information (privileged information) alongside visual features from fundus images.
• To address the scenario where genetic data is available for training but not for testing.

Main Methods:

• Utilizing the learning using privileged information (LUPI) paradigm.
• Training a predictive model that incorporates SNPs as privileged information to enhance fundus image-based glaucoma detection.
• Developing a model to predict glaucoma using visual features from fundus images, informed by genetic data.

Main Results:

• Demonstrated the effectiveness of the proposed LUPI approach in glaucoma detection.
• Showcased the utility of incorporating genetic information (SNPs) to improve the performance of models relying solely on fundus images.
• Validated the approach through extensive experiments, confirming its usefulness in a practical setting.

Conclusions:

• The proposed LUPI-based method effectively integrates genetic information for enhanced glaucoma detection using fundus images.
• This approach offers a valuable solution for leveraging privileged genetic data in machine learning models when it's unavailable during testing.
• The study highlights the synergistic potential of combining genetic markers and imaging data in ophthalmological diagnostics.