Optic nerve head assessment in Light-field fundus images – A case study

Abstract Purpose A suspicious optic nerve head (ONH) appearance, with abnormal cupping or increased cup-to-disc ratio, is associated to a strong risk factor for glaucoma. Typical devices to analyse the structure of the ONH are optical coherence tomography or scanning laser ophthalmoscopy. Those devices provide parameters to describe the status of an ONH like its depth, width or volume. We show that these parameters for glaucoma diagnosis can be obtained with a commercial fundus camera using an adapted light field (LF) camera providing a one-shot 3d retinal image. Methods We examined a human eye (32 years, male) 5 times during one session in compliance with the Declaration of Helsinki. The images were taken with a fundus camera (FF450, Carl Zeiss Meditec AG, Germany) combined with an industrial LF camera (RX12, Raytrix GmbH, Germany). The generic depth estimation was performed with the Raytrix LF software (RxLive 5.0.031.0) using the Depth Raycast Light algorithm. The results were exported as 3d point lists in ascii XYZ format and for further processing imported in Matlab (R2018b, MathWorks, U.S.). The imported data were pre-processed, the papilla midpoints automatically detected and the optic nerve head was horizontally sliced in 0.01 virtual depth steps starting from the base upwards until the papilla rim was reached. We compared the overall virtual depth, the virtual volume and the virtual area per virtual depth step, each as relative difference. Results All LF images showed visually a very good image quality. The depth estimations were successful and covered the complete image areas. The overall virtual depth range was 95.9% to 101.4% with mean of 100% (±2.36%). The overall virtual volume range was 97.5% to 102.2% with mean of 100% (±2.1%). The area per virtual depth range was 2.5% to 174.5% with mean of 100% (±10.9%) whereas the larger deviations appeared in the lowermost slices. Conclusions Fundus cameras with LF technique are able to provide one-shot 3d retinal images. This light field images allow the evaluation of typical parameters for glaucoma diagnosis and show a very good reproducibility.