No description available.

No description available.

No description available.

No description available.

No description available.

No description available.

Computers now pervade the field of medicine extensively; one recent innovation is the development of intelligent decision support systems for inexperienced or non-specialist pbysicians, or in some cases for use by patients. In this thesis a critical review of computer systems in medicine, with special reference to decision support systems, is followed by a detailed description of the development and evaluation of two new, interacting, intelligent decision support systems in the domain of diabetes. Since the discovery of insulin in 1922, insulin replacement therapy for the treatment of diabetes mellitus bas evolved into a complex process; there are many different formulations of insulin and much more information about the factors which affect patient management (e.g. diet, exercise and progression of complications) are recognised. Physicians have to decide on the most appropriate anti-diabetic therapy to prescribe to their patients. Insulin-treated patients also have to monitor their blood glucose and decide how much insulin to inject and when to inject it. In order to help patients determine the most appropriate dose of insulin to take, a simple-to-use, hand-held decision support system has been developed. Algorithms for insulin adjustment have been elicited and combined with general rules of therapy to offer advice for every dose. The utility of the system has been evaluated by clinical trials and simulation studies. In order to aid physician management, a clinic-based decision support system has also been developed. The system provides wide-ranging advice on all aspects of diabetes care and advises an appropriate therapy regimen according to individual patient circumstances. Decisions advised by the pbysician-related system have been evaluated by a panel of expert physicians and the system has undergone informal primary evaluation within the clinic setting. An interesting aspect of both systems is their ability to provide advice even in cases where information is lacking or uncertain.

Various methods are investigated for selecting training data for the purpose of training neural networks. A new method called MIQR (Maximum Inter-Quartile Range) is proposed for effectively selecting a concise set of training data. In addition, the ensemble concept is introduced in this new method. Data selection is not unduly influenced by “outliers”, rather, it is principally dependent upon the “mainstream ” output of the ensemble networks. Encompassed in the new method is a very simple ancient Chinese philosophical idea, i.e. “the minority obeys the majority”. These techniques are nonparametric in the sense that several different neural networks comprise an ensemble or committee and co-operatively work together with each other to achieve a common goal. Because these are different neural networks (hybrid model), they can be complementary in the entire learning system, and therefore effectively enhance the entire learning system’s efficiency and accuracy. For learning, the neural networks attempt to actively select the most informative and important training data. The methods described in this thesis pleasingly satisfy this need, and compare favourably with contending methods. Many experiments have been done to corroborate theoretical and empirical conjectures. The results are quite pleasing in that this new method is not only as “active learning” much better than “passive learning” both in data selection and in generalisation performance, but also outperforms other existing contending active learning methods. In particular, the results are very satisfying and interesting when the method is applied to discontinuous functions. Although the experiments are conducted with clean data selection, it should be easy to extend them to noisy data selection since the method developed is validated using unlabelled data. The algorithm developed for these methods has been rigorously tested, and proves to be highly autonomous and robust. The methods developed here are not restricted to use on neural networks. More generally, they can be applied to other scientific research and economic fields, even educational and sociological behaviour.

Machine Learning techniques, in particular induction algorithms, have been applied to the field of expert systems development in an effort to overcome the knowledge acquisition bottleneck. Many different induction algorithms have been developed. These utilise a number of different knowledge representations: e.g decision trees and rules. The rule based representation includes systems which utilise both propositional and predicate logic languages. Decision tree and rule based induction employ different knowledge acquisition mechanisms, however both strategies tend to induce complex and inaccurate knowledge from problem domains that contain noise or representational complexity. Previous techniques have concentrated on changing either the knowledge acquisition mechanism or the final ruleset/decision tree to reduce complexity. This thesis presents a new approach that focuses induction on those members of a training set that are likely to provide reliable knowledge. This is achieved by a new measure which is used to identify the most representative examples in a training set. The results of experiments show that this approach produces much simpler rulesets which in some circumstances perform with greater accuracy on unseen data.

The aim of the thesis is to evaluate recent work in artificial intelligence (AI). It is argued that such evaluation can be philosophically interesting, and examples are given of areas of the philosophy of AI where insufficient concentration on the actual results of AI has led to missed opportunities for the two disciplines — philosophy and AI — to benefit from cross-fertilization. The particular topic of the thesis is the use of AI techniques in psychological explanation. The claim is that such techniques can be of value in psychology, and the strategy of proof is to exhibit an area where this is the case. The field of model-based knowledge-based system (KBS) development is outlined; a type of model called a conceptual model will be shown to be psychologically explanatory of the expertise that it models. A group of major philosophies of explanation are examined, and it is discovered that such philosophies are too restrictive and prescriptive to be of much value in evaluating many areas of science; they fail to apply to scientific explanation generally. The importance of having sympathetic yardsticks for the evaluation of explanatory practices in arbitrary fields is defended, and a series of such yardsticks is suggested. The practice of providing information processing models in psychology is discussed. A particular type of model, a psychological competence model, is defined, and its use in psychological explanation defended. It is then shown that conceptual models used in model-based KBS development are psychological competence models. It follows therefore that such models are explanatory of the expertise they model. Furthermore, since KBSs developed using conceptual models share many structural characteristics with their conceptual models, it follows that a limited class of those systems are also explanatory of expertise. This constitutes an existence proof that computational processes can be explanatory of mental processes.