Abstract
Abstract: :
Purpose: Fundus cameras are used to view and document retinal disease. Digital photography through fundus cameras offers new opportunities for computer–assisted detection and quantification of pathology. However, intrinsic fundus camera characteristics introduce variables in retinal photography that may compromise analysis accuracy. One such characteristic, lens distortion, is often overlooked in traditional CCD camera or computer color calibration regimens. Additionally, the degree of variability among fundus cameras of the same make and model is unknown. We evaluated lens distortion among two fundus camera models from the same manufacturer. Methods: Calibration targets were created and photographed using six fundus cameras: three TopconTRC–50EX mydriatic and three Topcon TRC–NW6S nonmydriatic fundus cameras. Algorithms were used to undistort captured images and measurements performed. Mydriatic images were captured withh 20, 35, and 50–degree fields. Nonmydriatic images were obtained using 30 and 45–degree fields. Results: We estimated three nonlinear terms (k1, k2, k3) of lens radial distortion (pincushion and barrel) and found measurements over or underestimated by varying amounts depending on measurement location, camera and field of view. In one example, radial distortion in a mydriatic 50–degree field image was found to be k1= –0.43, k2= –0.22 and k3= 0.00, resulting in a 26% distortion factor. Conclusions:Accurate measurement of eye disease pathology in digital imagery requires precise calibration of fundus camera operating characteristics. Accounting for lens distortion is an important component of any fundus camera calibration regimen.
Keywords: imaging/image analysis: clinical • imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • retina