Purchase this article with an account.
L. V. Raju, M. Palaiou, T. R. Friberg; Comparison of Computerized Detection and Measurement of Drusen Between Trained and Novice Observers. Invest. Ophthalmol. Vis. Sci. 2007;48(13):5112.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To compare the variability in detection and measurement of drusen between trained and novice observers using the Drusen AnalyzerTM (Iridex, Mountain View, CA) program.
Fifty 35 mm Kodak Ektachrome slides which had been taken at baseline for the Age-Related Eye Disease Study (AREDS) at the University of Pittsburgh had been previously read by the AREDS Reading Center and were available for our study. Informed consent was obtained which adhered to IRB and HIPAA guidelines. Drusen readings within two concentric fields centered at the fovea (1000 and 3000 microns diameter) were performed by a novice reader (LVR) on three separate days and also by a retinal specialist with many years of clinical experience and in using the Drusen AnalyzerTM program. The novice observer, a second year ophthalmology resident, was given a 30 minute orientation to the program and no specific direction as to what had been deemed a druse. The drusen measurements obtained by the novice were then compared to a trained observer’s measurements, which were taken as the standard.
If the analyzer was easy to learn and accurately identified drusen in color images, we would expect to find a good correlation between the readings of the trained and novice reader. Our results show that over time the drusen measurements by the novice became more conservative and had a closer correlation to the standard measurements, evidenced by correlation values of 0.65, 0.72, and 0.79 for the first through third trials respectively. The correlation between the novice’s first and second trials (as a measure of precision) and the second and third trials increased from 0.77 to 0.85. This suggests that there was a process of learning involved when using the different aspects of the program. A paired T-test also showed that measurements of the novice observer were statistically similar to the trained observer’s with a p value <0.001 for both the 1000 micron and 3000 micron areas of regard.
Our studies show that use of the Drusen AnalyzerTM program would be helpful in the quantification of drusen by observers of different experience levels. While the process of drusen analysis now can be somewhat time intensive and prone to subjectivity, the use of such a program may be helpful in creating a baseline level of objectivity. The results show that the data is reproducible and with a brief orientation to its nuances, the program can be accurately employed.
This PDF is available to Subscribers Only