Medical Decision Making: A Case Study Within the Cardiology Domain

摘要: Objective: Optimal patient management is dependent on correct diagnosis and decisions. However decision making in medicine a complex challenging problem often inaccurate. This can be illustrated by considering the of diagnosing acute myocardial infarction (AMI). Studies indicate that as many 20% patients with AMI are wrongly diagnosed. Computerised medical support systems have potential to improve this accuracy. In paper we evaluate performance Kernel based Support Vector Machine (KerSVM) learning methods which new, efficient effective methodology for solving pattern classification problems. Methods: Using framework generalised model, tested validated behaviour five kernel classifiers: Linear, Polynomial, Gaussian, Laplacian Sigmoid respect AMI. It was data taken from 242 who had presented chest pain; 130 these correctly been diagnosed suffering other 112 their pain attributed causes. Initially model using 105 individual pieces history, electrocardiography, echocardiography blood investigations. Results: The diagnostic accuracy SVM different kernels varied between 86.7% 93.8%. also run through 3 well-established classifiers -Bayesian, Decision Trees Neural Networks, results compared favourably those model. We carried out further experiments several subsets features. found 68 features (taken tests) or 66 electrocardiography echocardiography) it possible obtain accuracies 90% 93% classifiers. Conclusion: overall demonstrate effectiveness robustness proposed approaches assisting clinicians accurate making.