AV-RetVasc: an ultra-wide-field retinal vascular segmentation dataset from real-world retinal vascular disease scenarios

However, existing retinal vessel segmentation datasets are mainly derived from conventional color fundus photographs of healthy populations or eyes without prominent vascular abnormalities, limiting their applicability to real-world pathological conditions.

Ultra-wide-field (UWF) imaging provides a broader retinal view of approximately 200 degrees and enables comprehensive visualization of peripheral vascular abnormalities, but pathological UWF vessel segmentation remains challenging due to complex vascular morphology, vessel scale variation, image distortion, and interference from retinal lesions. Here, we present AV-RetVasc, an ultra-wide-field retinal vessel segmentation dataset derived from patients with retinal vascular diseases.