This PhD project funded by SEED ( Software Engineering Experience Development ) sponsors M. Mostafizur Rahman Mail Mr M.M.Rahman AT 2009.hull.ac.uk
As medicine moves forward there is a need for sophisticated decision support systems that make use of data mining. Existing models, usually based on linear statistical analysis, have proved disappointing. Statistical methods such as statistical regression, Cox proportional-hazards regression, logistic regression, inverse variance weighted method; or groups' comparison have been commonly used in different studies (Bellamy et al., 2007; Howard et al., 2006, Ruijter et al., 2009; Wang et al., 2006). However, these methods are usually used to explain the data and to model the progression of the disease rather than to make predictions for populations or individual patients. The adoption of clinical governance in the NHS has mandated that we must develop appropriate and reliable clinical data-sets for use in comparative audit. These data-sets will be useless without the ability to interrogate and analyse them in a meaningful way. A validated model would allow us to set achievable national standards and thereby to improve quality of care throughout vascular units in the UK by implementing guidelines and allowing comparative audit using local and national data-sets. This project investigates a systematic methodological approach to developing decision support systems for clinical domains using fuzzy logic throughout the process. We aim to show that such systems are capable of meeting the above requirements. We may also use hybrid models based on different AI techniques to develop decision support systems whose rules can be extracted and analysed by a human being. The bulk of the research will be in finding the suitable data preparation technique for feature cleaning and reduction and developing fuzzy or fuzzy neural techniques suitable for inclusion in such systems, and the use of fuzzy logic in combing classifier outcomes for practical decision making in medicine. Clinical data from cardiovascular medicine and other domains (Merz & Merphy, 1990) is available for use in this project.