Classification of heart conditions based on heart sounds

This paper discusses the classification of paediatric heart problems by examination of heart murmurs, which are flow noise in addition to the two main heart sounds. Each type of heart murmur has unique acoustic characteristic, which. enable doctors to obtain valuable information about the condition of the heart. In this paper, these characteristics are quantified and their distributions compared. Features are extracted from their time-frequency representations and are used to train the self-organising maps, a type of unsupervised artificial neural networks. On the resultant two dimensional self-organised maps, samples from four groups of patients fall into four clusters showing the four groups have inherently different distributions.