Digital imaging in diabetic retinopathy screening - a comparative study of two digital cameras

The National Screening Committee (NSC) has recommended digital imaging for diabetic retinopathy screening. There have been recommendations but no consensus yet on the specifications for the digital cameras to be used for the purposes of diabetic retinopathy screening. We therefore undertook a study to evaluate two different digital cameras with different image resolutions cur- rently being used in UK for diabetic retinopathy screening to detect any difference in their evaluation of retinal details. Methods. The study involved detailed assessment of 192 images taken from 48 diabetic patients recruited from two centres - City Hospital, Birmingham, and Llandough Hospital, Cardiff. Digital retinal photographs of both eyes were obtained using the Sony DXC-950P and the JVC KY-F70U mounted on a Canon CR6-45NMf non-mydriatic retinal camera. The JVC camera has an image resolution of 1360 × 1024 pixels similar to the NSC guidelines and produces as a consequence an image with a 4 MB file size. The Sony on the other hand has an image resolution of 750 × 570 pixels and generates a file size of only 1.3 MB. The images were cap- tured using 'EyeCap' software (Orion Imaging, UK) on a Pentium III PC and displayed on a Sony Trinitron colour monitor. Comparison was then made between the JVC and Sony images in bitmap (BMP) formats. Results. 46 patients were included in the study, of whom 27 were male and 19 female subjects. The median age was 60 years (range 27-89). They were from a varied ethnic background and the majority had type 2 diabetes (n = 43). A total of 184 digital images were available for evaluation. 17 images could not be graded because of cataracts (15) and asteroid hyalosis (2). The difference in retinopathy grades between the two digital cameras were not found to be statistically significant (p = 0.943). Overall agreements in grading between the two cameras were seen in 85/92 (92.39%) images. Conclusion. For the purpose of screening for diabetic retinopathy, both retinal cameras are comparable, and one does not appear to have any major advantages over the other. This may also suggest that the resolution of neither camera is optimum and perhaps much higher resolution digital cameras may be needed for diabetic retinopathy screening in the future. File sizes are an important issue with regard to storage of images as well as transmission over computer networks. This may become a difficult issue, as higher reso- lution cameras will inevitably require larger amounts of memory for storage. Currently there appears no justification in using the JVC camera with larger file sizes instead of the lower resolution Sony DXC-950P camera. Copyright © 2002 John Wiley & Sons, Ltd. Practical Diabetes Int 2002; 19(9): 287-289