Purpose:
To validate a new computer assisted grading tool for evaluating Age-Related Maculopaty (ARM) features using digital color fundus images.
Methods:
Twenty-five RE/LE 45º digital stereoscopic colour fundus images were acquired from 48 participants, using a Topcon TRC- 50Dx mydriatic camera. Grading of ARM features was then performed by 3 senior ophthalmologists using the RetmarkerAMD (Critical Health SA), a computer assisted grading system that follows the International Classification Grading System guidelines, to check for agreement and validate the system. With the RetmarkerAMD, graders were able to pinpoint (draw-over) significant ARM features such as: Drusens using 5 circles with standard dimensions (C0: <63µm; C1: 63µm<125µm; C2: 125µm<175µm; C3: 175µm<250µm; C4: 250µm<500µm); Hyper and Hypopigmentation using standard size circles to document these alterations; Geographic Atrophy and Exudative Lesions using freehand drawing tools. For each lesion type the system used a unique color-code representation. The system overlay a standard grid dividing the eye fundus image in 10 sub-fields for graders geographic reference: Field 1 (central circle centered in fovea); Fields 2 to 8 (inner circles); Field 10 (outer circle). The system calculated automatically lesion number and area. The results were displayed onscreen in real time. The agreement between graders was tested using the kappa coefficient.
Results:
A good agreement was found for drusen greater or equal to 125µm, k=0.622 (95% CI=[0.438; 0.806]), for the number of drusen, k=0.753 (95% CI=[0.586; 0.920]) and for the total area occupied by the drusens. A fair agreement was found for the number of hard drusen (k=0.34) and a weaker agreement was found for the quantification of hyper and hypopigmentation (k=0.11 and k=0.06, respectively). The new software was found to be more effective, less time consuming (near 35% faster), allowing the identification of more lesions (near 32% than standard manual grading system) and the reduction of human error.
Conclusions:
The RetmarkerAMD showed a good agreement between graders and reveals to be a less time consuming and more effective tool for ARM lesions detection. This system allows the quantification and localization of the different AMD features in the different sub-fields. It can be used as a platform for clinical trials and epidemiologic studies and is being used in an ongoing epidemiologic study for ARM in Portuguese population with an estimated number of 4000 participants.
Keywords: age-related macular degeneration • clinical (human) or epidemiologic studies: systems/equipment/techniques • clinical (human) or epidemiologic studies: prevalence/incidence