Human observers are subjective and prone to inter-observer and intra-observer variability;
thus three independent observer marks are provided, each of whom marks a set of
randomly selected vascular segments.
Observer 1
The first observer (O1) is an ophthalmic specialist with research interests in automated
assessment of diabetic retinopathy and evaluation of retinal vascular geometry;
he has five year experience in retinal fundus images analysis.
Observer 2
The second observer (O2) is a specialist optometrist and a qualified secondary grader
for diabetic retinopathy; he has been previously involved with different research
projects involving handling and analysis of digital images with tasks for defining
vascular and non vascular diabetic lesions.
Observer 3
The third observer (O3) has a satisfactory basic medical background being an experienced
laboratory specialist. She received a training session in defining and marking retinal
vascular segments before starting independent marking of the selected segments for
the experiment. Her initial training set was reviewed and independently assessed
by an ophthalmic consultant.
The Reference Standered
The mean of the measured diameters from the three observers is used as the reference
standard (O).