Page 106 - Computational Retinal Image Analysis
P. 106
2 Benchmark datasets and evaluation metrics 99
AVR ✓
Tortuosity ✓
Junctions ✓
Table 1 A summary of representative retinal image datasets and benchmarks.
A/V ✓ ✓ ✓ ✓ ✓ ✓
Vessel trees a ✓ ✓ ✓ ✓ ✓ Notes: For retinal imaging technique (IT), CF refers to color fundus photography, SLO is for scanning laser ophthalmoscopy, FA is for fluorescein angiogram, and UWFI is for ultra-wide fundus imaging. “Vessel masks” denotes vessel segmentation annotation, while “Vessel trees” is for annotation of individual vessel trees, and “A/V” stands for ar
Vessel masks ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ a For DRIVE, IOSTAR, VICAVR, and INSPIRE-AVR, individual vessel trees out of the OD are manually annotated and make available online (http://imed.nimte.
No. of images 40 20 45 28 143 40 58 40 60 24 30 8
IT CF CF CF CF CF CF CF CF CF SLO UWFI FA readouts (see Section 4.4 for details). ac.cn/vetovessel-topology-groundtruth.html).
Database DRIVE [24] STARE [25] HRF [29] CHASEDB1 [30] ARIA [31] DR HAGIS [32] VICAVR INSPIRE-AVR [33] RVTD [34] IOSTAR [35] AV-WIDE [36] VAMPIRE [37]
