Effect of Fundus Camera’s Lens Distortion on Digital Fundus Images

Author: Adol Esquivel, MD

Primary Advisor: Cynthia Phelps, PhD

Committee Members: Jack W. Smith, MD, PhD

Masters thesis, The University of Texas Health Science Center School of Health Information Sciences at Houston.

 
PURPOSE. Advances in technology are leading to increasing use of optical fundus cameras equipped with digital sensors. Digital imagery offers the potential of computer-based analysis to measure, classify, and track the progression of retinal pathology. In this study six digital fundus cameras were examined to determine degrees of distortion introduced by their optical systems.
METHODS. The optical systems of six fundus cameras (3 mydriatic and 3 non-mydriatic) were calibrated using two physical targets to determine each camera system’s intrinsic and extrinsic parameters. Target A was a 2” x 2” 2D checkerboard mounted on a tripod and Target B was a 60” x 60” 2D checkerboard mounted on a wooden frame. Algorithms were then used to compensate digital images for lens distortion and principal point deviation using a ring diagram. Area changes in the ring diagram due to lens distortion were calculated. Variability of image areas due to image’s mask within the cameras’ study group was also compared.
RESULTS. The variability of imaged areas inside the mask of mydriatic 20°, 35° and 50° lenses ranged from 0.1% - 1.7%. Non-mydriactic camera masks differed by 2% for the 30° FOV and 11% for the 45°. Variable lens distortion was observed among the cameras and among fields of view. Target A principal points showed deviation in the range of 2% - 8% for mydriatic cameras and 1% - 2% for non-mydriatic cameras. Analysis of Target B principal points showed deviation in the range of 1% - 10% for mydriatic cameras and 0% - 2% for non-mydriatic cameras. Area change was highest at the most peripheral rings and the widest fields of view.
CONCLUSIONS. the study suggests that different optical fundus cameras have different intrinsic properties, even when they are identical model high-quality instruments made by the same manufacturer. It was shown that lens distortion affects images obtained using fundus cameras with wide fields of view. Fundus camera variability as evidenced by our results makes a strong case for calibration.