Data Integrity and Intelligent Data Analysis Techniques Applied to a Glaucoma Progression Dataset

Glaucoma is a condition that affects the human eye and is an umbrella term for a family of related eye conditions. A common trait of these conditions is a functional abnormality of the retina and optic nerve, leading to loss of visual field. This vision loss is usually only part of the visual field, although, untreated, glaucoma often leads to blindness. It is thought that by 2010, there will be over 60 million people worldwide suffering from the various forms of Glaucoma. Visual field tests are crucial to the diagnosis and management of all types of glaucoma. Such tests require the level of retinal sensitivity to light to be sampled at a number of points typically between 50 and 100, depending on the type of test; the eye is then assigned a numerical value in the range 'no perception' (lowest) to 'perfect perception' (highest). A specialised machine is used to conduct these tests - a typical clinical test can take between six and seven minutes per eye. Once diagnosed with glaucoma or suspected glaucoma, a patient is monitored and recommended to undergo the same tests every six months (or more frequently in some cases). However due to the psychophysical nature of the glaucoma test, the results can therefore vary in quality dramatically. For example, they can be affected by patient fatigue (as the test can last for long periods) and attention span deficits (particularly in the elderly and children).This proposal aims to use data quality metrics (such as false positive and negative rates) to incorporate uncertainty into computational models that will also take into account the spatial and temporal nature of visual field data. The proposed research will use probabilistic Cellular Automata (CA) with an appropriate rule learning approach as the technique to model the visualfield deterioration of glaucoma sufferes. The aim is to accurately model visual field progression and to provide an aid to the clinical practitioners.This project is in collaboration with Moorfields Eye Hospital, London, UK.

This project will have an impact on two major groups. First, our research benefits the group of clinicians in the front line of glaucoma treatment and support. The results of the project will show how advanced computation techniques can be used in the analysis and diagnosis of glaucoma. In many cases, clinicians would have used conventional statistical methods in their modelling, and have rarely used the novel techniques that we propose. By demonstrating the value and worth of methods such as cellular automata and genetic algorithms, we will open up a new frontier of potential research opportunities that has not yet been considered, explored and exploited. We will ensure that they have the opportunity to benefit from our research project by disseminating our work to those clinicians through our proposed workshop plans, and using this platform to champion the benefits of these advanced techniques. We also intend to use the workshop and our industrial links as opportunities for forging new research collaborations and disseminating our research in other ophthalmological areas. Second, our research will aid patients suffering from glaucoma. Through the more accurate prediction and modelling of glaucoma deterioration, we will be able to vastly improve the level of patient treatment. Significantly greater modelling accuracy leads directly to more effective patient treatment and an improvement in the quality of lives of those patients. The quality of the methodology of the proposed research means that tangible and quantifiable improvements will be made to patient treatment.