Analysis of Heart Sounds using Hidden Markovian Models

Auscultation has long been the primary test for initial assessment of the patient's heart conditions and early detection of anomalies. However, the reliability of this technique on subjective judgement, expertise, and individual's hearing accounts for the inconsistency in the diagnostic among experts. Despite several efforts towards the development of autonomous systems, their limited success suggests the use of new approaches for signal representation and pattern classifiers. This book explores the use of hidden Markovian models (HMM) for characterisation and classification of heart sounds. The best results, obtained using cepstral representations, are compared to those provided by paediatric cardiologist's diagnosis based on X-ray plates and electrocardiograpy. Both results are combined in a sequential tests to assess the use of the HMM classifier as a clinical diagnosis aid. Signal processing practitioners, researchers and students in the fields of biomedical signal processing, statistics, computer sciences, engineering and medicine will benefit from this work.

Autorentext

PhD in signal processing from the Institute of Sound and Vibration Research of the University of Southampton, UK. Researcher at the Northwest Biological Research Centre (CIBNOR), working in Engineering applied to Biology. He is a member of the Signal Processing, and Engineering in Medicine and Biology IEEE Societies