Automated Image Analysis of Postertior Capsule Opacification

Research project P 14269 Automated Image Analysis of PCO Oliver FINDL, Axel PINZ, Rupert MENAPACE 06.03.2000 Background: PCD, or after-cataract, is the most long-term complication of modern cataract surgery with an incidence of about 25% within 5 years. The treatment for PCO is to cut an opening into the posterior capsule with a neodymium: YAG laser. However, laser treatment carries the risk of additional complications, is costly and often not available, especially in developing countries. For this reason a large number of clinical trials are underway to assess the efficacy of surgical techniques, modifications in IOL design and material, and drugs to reduce the occurrence of PCO. In order to analyse PCO, the capsule needs to be photographed with the retrolillumination technique. Since this causes disturbing light reflex artifacts, especially in the centre of the image, the most important zone for visual function cannot be analysed. To date there is no objective system for quantification for the essential and main outcome variable, namely PCO intensity. However, objective quantification is mandatory for assessment of PCO intensity in prospective, randomised, clinical trials. Purpose: To develop a reprocucible photographic system and fully automated software system for objective quantification of PCO. Rationale: Objective and quantitative direct measurement of lens epithelial cell proliferation on the posterior capsule would allow for more accurate evaluation of PCO. This would increase the sensitivity for measuring small changes in PCO during shorter follow-up periods and increase discrimination for small differences between groups in prospective studies. Therefore, the number of patients needed for comparative studies will decrease which has ethical impact. Method: Development of an algorithm to improve reflex artifacts from the photos. Segmentation of PCO areas will be attempted with the use of the local co-occurrence matrix method. A learning set of 300 images of PCO will be graded by experienced examiners and defined as the `ground truth`. Possibly a combination (fusion) of the results of several algorithms will be needed to optimise the analysis system. In a second step, the algorithms will be operated by clinicians and thus be tested in clinical use. A test set, consisting of longitudinal data from clinical trials, will be used to assess the analysis systems.

Posterior capsule opacification (PCO, or after-cataract) is the most common long-term complication of modern cataract surgery with implantation of an intraocular lens (IOL). It is caused by lens epithelial cells that migrate and grow between the posterior surface of the IOL and the posterior lens capsule. PCO can be treated by cutting an opening into the posterior capsule with a Nd:YAG laser (YAG laser capsulotomy). However, this procedure causes additional costs and can lead to new complications. Therefore, many studies deal with different ways of inhibiting PCO formation. In order to compare different clinical trials, an objective and standardized method of PCO measurement is needed. The aim of this cooperative project between the Department of Ophthalmology, University of Vienna, and the Institute of Electrical Measurement and Measurement Signal Processing, Graz University, was to develop a reliable, time-sparing, easy-to-use method of assessing after-cataract that includes no subjective step. The product of our research work is a prototype version of the AQUA software ("Automated Quantification of After-Cataract"). This system is to be used for prospective, randomized, masked clinical trials assessing the effectiveness of surgical techniques, drugs, IOL design and IOL material on preventing PCO. The AQUA system is based on a digital image acquisition system for acquiring high quality digital retroillumination images of the posterior capsule and a PC-based program that calculates the entropy (grade of disorder) of the region of interest in the digital image. The entropy value is then converted to the "AQUA score", ranging from 0 to 10, which was shown to correlate well with subjective scoring of PCO images by experienced examiners. The software was already used in several clinical trials comparing different IOL designs. We could show that an IOL with a sharp posterior optic edge design significantly reduces PCO formation compared to IOLs with round optic edges. Our image data set at the Department of Ophthalmology meanwhile consists of several thousand PCO images which are still being evaluated. Currently, the AQUA system calculates a global PCO score (based on the entropy value) for the entire region of interest (the posterior capsule). We are currently working on an improved PCO assessment method with fully automated segmentation of a digital PCO image into differently textured areas which should lead to even more accurate results and will be less sensitive to image artifacts caused by the flash light of the image acquisition system or by structures that reside outside the plane of sharp focus. It should also work with digital images from other camera systems and conventional photo slit lamps that are commonly available. The aim is to develop a final version of this improved AQUA algorithm during a new cooperative project between the University of Vienna and the Technical University of Graz.