Assessment of Posterior Capsule Opacification after Cataract Surgery

Posterior capsule opacification (PCO) is the most common long-term complication with more than 25% overall incidence within 5 years after cataract surgery of the human eye. The goal of the project described in this paper is to devise a fully automated, objective measure of grades of PCO from so-called retroillumination photographs. We review the current state of the art of existing PCO grading systems, which are either manual or semi-automated and mostly rely on the segmentation / interpretation / grading of textures. We present several algorithmic improvements as well as a new approach to obtain reflection-free images by the fusion of several photographs. A comparative experimental study applying three different grading systems to images of more than 100 cases shows very good results for a prototype implementation of our new PCO grading system AQUA (Automated QUantification of After-cataract ). keywords: texture segmentation, classification, fusion, PCO, automated grading 1. POSTERIOR CAPSULE OPACIFICATION Cataract surgery has become a standard process in eye surgery. Compared to the formerly used intracapsular cataract extraction (ICCE), the extracapsular cataract extraction (ECCE) decreases the risk of complications. During this operation the opaque lens is replaced by an artificial one, the intraocular lens (IOL). The most frequent postoperative complication associated with decreased vision is the posterior capsule opacification (PCO) caused by lens epithelium cells (LEC) growing between the posterior capsule and the IOL (see Fig. 1). The PCO can be treated by YAG laser capsulotomy which possibly, though infrequently, leads to new complications such as retinal detachment, endophthalmitis or intraocular pressure rise. But these new complications may even annul the accommodation that is attained by IOL implantation [13]. Additionally, laser treatment poses a cost factor for the health care system which should not be underestimated; e.g. Medicare spent over 250 million dollars for laser capsulotomy in 1993 in the USA [2]. In fact, laser capsulotomy already is the second frequently performed ophthalmic surgical procedure after cataract surgery in the USA. These enormous costs explain why ECCE with IOL implantation is not extensively performed in the Third World. Therefore numerous efforts are made to fight PCO. Strategies to prevent PCO are e.g. the surgical removal of LEC, toxic inhibition or the inhibition of LEC migration. But LEC proliferation also depends on IOL material and design. It becomes clear that a way of measuring LEC proliferation is necessary to assess the efficiency of these strategies. In most clinical trials the laser capsulotomy rate has been used as the main outcome measure [12, 17]. Unfortunately, the capsulotomy rate mostly depends on the clinician who performs the assessment of the PCO. Other factors which affect these studies are patient demands, financial situation and access to laser equipment. There are some attempts to objectify the assessments of the clinicians by computer evaluation of areas marked by the clinicians. But these approaches like LOCS, OPAC, GRID [20] or EPCO [19] in fact still depend on the assessment of the clinician and thus are subjective. Obviously, only an automated quantification of PCO grants objective assessment in prospective, randomized and double-blinded clinical trials. 2. AUTOMATED GRADING 2.1. Recording Techniques Various studies deal with measuring density of opacification in a Scheimpflug image [11, 9] which uses the correlation of opacification density and visual acuity under conditions of PCO. Since the method uses cross-sectional images of the IOL and the posterior capsule, no morphological data of PCO is supplied. But in fact, it is the longitudinal analysis of PCO morphology which is an intrinsic part of the assessment of PCO to find out the most effective ways of preventing PCO. A morphological scoring system of standardized retroilAnterior Rhexis Cornea Haptics