Ophthalmic Biomarker Detection: Highlights from the IEEE Video and Image Processing Cup 2023 Student Competition

TL;DR

This paper discusses a student competition focused on developing AI methods for detecting biomarkers in 3D OCT images, emphasizing the challenge of balancing personalization and generalization across diverse patient data.

Contribution

It introduces a real-world biomedical imaging challenge with a multimodal OCT dataset, highlighting the importance of addressing domain variability in AI biomarker detection.

Findings

Students developed methods to handle domain differences in OCT scans

The competition showcased approaches to balance personalization and generalization

The dataset facilitated research on AI in biomedical imaging

Abstract

The VIP Cup offers a unique experience to undergraduates, allowing students to work together to solve challenging, real-world problems with video and image processing techniques. In this iteration of the VIP Cup, we challenged students to balance personalization and generalization when performing biomarker detection in 3D optical coherence tomography (OCT) images. Balancing personalization and generalization is an important challenge to tackle, as the variation within OCT scans of patients between visits can be minimal while the difference in manifestation of the same disease across different patients may be substantial. The domain difference between OCT scans can arise due to pathology manifestation across patients, clinical labels, and the visit along the treatment process when the scan is taken. Hence, we provided a multimodal OCT dataset to allow teams to effectively target this challenge. Overall, this competition gave undergraduates an opportunity to learn about how artificial intelligence can be a powerful tool for the medical field, as well as the unique challenges one faces when applying machine learning to biomedical data.