8.1 Introduction

At some time in their lives 60 80 per cent of the population will experience one episode of low-back pain (LBP), of whom 90 per cent will get better within six to eight weeks without need for treatment or investigation [1,2]. The remaining ten per cent incur 70 90 per cent of the medical costs arising from low-back pain and represent a challenge to health practitioners in providing an accurate diagnosis and successful management [3]. The number of patients who receive a specific diagnosis is small and several authors have highlighted the difficulty of diagnosing LBP [3,4]. Low-back pain is a multifactorial problem that includes physical, psychological and social aspects of illness [2].

The costs of back pain to society are very high and continue to rise. UK government statistics [1] for 1993 94 estimate 106 million working days lost in the UK due to back-pain incapacity, 480 million annual National Health Service costs for back-pain treatment, 3.5 billion costs to industry in lost production and 1.4 billion in social security benefits.

In previous research Bounds et al. [5,6] evaluated the performance of artificial neural networks (NNs) for the classification of low-back-pain patients. It was found that the mean diagnostic accuracy of the NN predictions was higher than the mean accuracy produced by the clinicians. Neural networks are being increasingly used as decision-support tools [7,8]...