Competition Introduction

Currently, development of automatic algorithms for ophthalmic imaging interpretation are faced with huge challenge due to lack of qualified data. Inaccurate label and no fine annotation prevents researchers from training algorithms with high efficacy. iChallenge is set up to share high-quality labeled and annotated imaging data of ophthalmology, to enhance communication between different researchers (not only computer scientists but also clinicians) and to promote development of automatic algorithms in diagnosis and image segmentation.

Ophthalmology is a unique branch of clinical medicine. There are various kinds of imaging examinations of ophthalmology: fundus photo, optical coherence tomography, fundus fluorescein angiography, scanning laser ophthalmoscopy and so on. A definite diagnosis of ophthalmic disease needs combination of results from several different tests. In clinical practice, diagnosis and treatment strategy determination relies on evaluation of imaging data. However, interpretation of ocular imaging data needs a lot of experience and time. Therefore, automatic algorithms with comparable accuracy to humans are strongly needed. And artificial intelligence is expected to help ophthalmologists in reading the images.

Currently, fundus photo has been widely used in diagnosis of glaucoma and retinal diseases. Here we provide you with 3 datasets with fine annotations and labels, including iChallenge-GON, the dataset of fundus photos on glaucoma; iChallenge-AMD, the dataset of fundus photos on age-related macular degeneration and iChallenge-PM, the dataset of fundus photos on pathologic myopia.