Drusen Area Measurement in the PTAMD Study: A Semi-automated Method vs a Manual Technique

Abstract

Abstract: : Purpose: Total drusen area is a risk factor for the development of visual loss and choroidal neovascularization (CNV) from age-related macular degeneration (AMD). We compared two methods used to measure drusen area in eyes undergoing prophylactic 810 nm photocoagulation. Method: We compared drusen-area measurements made on baseline digitized images of one eye in two groups of 18 patients, who had AMD and had at least five drusen greater than 63 microns in size and a visual acuity of 20/63 or better in the eye studied, and a disciform scar in the fellow eye. One patient group had prophylactic 810 nm laser treatment to drusen [Prophylactic Treatment of AMD (PTAMD) Study] and did not develop a choroidal neovascular membrane in the eye within 24 months followup, while the other group also had laser but subsequently developed CNV in the studied eye. The first method involved viewing the images on a monitor and using Adobe Photoshop^(TM). Each drusen was manually encircled using a cursor and the given area of all drusen was determined by totaling the number of enclosed pixels. A vein near the nerve was used as a scale, allowing pixels to be converted to square millimeters. The second measurement method utilized a specially written computer algorithm (from A.R.T.T., Crete, Greece; Iridex, Mt. View, California) which semi-automatically detects drusen; the operator using this software can adjust the threshold of detection and then make final interactive adjustments using encircling tools. The average time to conduct each analysis was noted. Results: By the automated method, the drusen area was 1.80 + 1.02 mm² for the CNV group and 1.00 + 0.53 mm² for the non-CNV group. By the manual method, the mean total drusen area was 5.32 + 3.36 mm² for the CNV group and 3.34 + 2.2 mm² for the non-CNV group. The differences between CNV and non-CNV group eyes were statistically significant. The normalized average standard deviation of the mean reading was smaller using the automated system (53% vs. 63% of the resultant area value). The time required to measure all drusen in the image within a circle of 4.5 mm was typically 5-7 minutes using the automated system and 15-20 minutes using the manual method. Conclusion: Semi-automated drusen detection is quicker and easier to perform than a manual technique. Both methods indicated that drusen area at baseline was significantly and substantially greater in eyes which developed CNV compared to those which did not.

Keywords: age-related macular degeneration • drusen • imaging/image analysis: clinical