COph100: A comprehensive fundus image registration dataset from infants constituting the "RIDIRP" database

Area of Science:

• Ophthalmology
• Medical Imaging
• Computer Vision

Background:

• Retinal image registration is crucial for diagnosing and treating eye conditions.
• Existing datasets lack sufficient infant data and clinical challenge representation.
• High-quality adult retinal datasets are limited in image pairs and real-world complexities.

Purpose of the Study:

• Introduce COph100, a novel dataset for infant retinal image registration.
• Provide a challenging resource with diverse image quality for algorithm development.
• Facilitate robust comparison of retinal registration methods in pediatric ophthalmology.

Main Methods:

• Curated 491 image pairs from 100 infant eyes across 2-9 sessions from the RIDIRP database.
• Manually annotated ground truth points for corresponding images.
• Generated automatic vessel segmentation masks for each image.

Main Results:

• COph100 dataset established with diverse image quality and clinical relevance.
• Dataset assessed using state-of-the-art registration algorithms.
• Provided ground truth and segmentation masks for enhanced analysis.

Conclusions:

• COph100 enables robust benchmarking of retinal registration algorithms.
• Facilitates the study of disease progression in infants.
• Advances understanding of pediatric ophthalmic conditions through improved imaging analysis.