A comparison of neural network and other pattern recognition approaches to the diagnosis of low back disorders

Abstract This article describes the application of Multi Layer Perceptron (MLP) and Radial Basis Function (RBF) networks to the problem of diagnosing low back pain and sciatica. The performance of these methods is compared with that of three groups of doctors, with another computer program (Norris, 1986) based on fuzzy logic ideas, and with K Nearest Neighbour and distance from class mean classifiers. One of these is equivalent to a Bayesian classifier—which has been applied extensively to medical diagnosis. Although clinical data on larger numbers of patients would be necessary before this neural network approach could be fully validated, the initial results are very promising. For the diagnostic category that it is most crucial to get right (because medical intervention may be required urgently), the MLP produces correct classification more often than any of the three groups of doctors or the fuzzy logic system. The mean diagnostic accuracy over all possible categories is also higher than that of the doctors and is comparable to that of the fuzzy logic system.