Retinal vessel measurement: comparison between observer and computer driven methods.

A method of semi-automated image analysis for the measurement of retinal vessel diameters is described. This was compared with an observer-driven method for reproducibility and accuracy. The coefficient of variation for the data from the semi-automated method was 1.5-7.5% (depending on the vessel diameter) compared to 6-34% with the observer-driven method. The mean vessel diameters using the observer-driven method tended to be higher; however, this did not reach significance. The speed and low inter- and intra-observer variability for the semi-automated method make it a useful technique for measuring retinal blood vessel diameters. A larger variability was found between photographs taken at different times. This may be due to changes in retinal vessel diameter with changes in retinal perfusion pressure during the cardiac cycle.