Bowel-sound pattern analysis using wavelets and neural networks with application to long-term, unsupervised, gastrointestinal motility monitoring

This work focuses on the implementation of an autonomous system appropriate for long-term, unsupervised monitoring of bowel sounds, captured by means of abdominal surface vibrations. The autonomous intestinal motility analysis system (AIMAS) promises to deliver new potentials in gastrointestinal auscultation, towards the establishment of novel non-invasive methods for prolonged intestinal monitoring and diagnosis over functional disorders. The system was developed utilizing time–frequency features and wavelet-adapted parameters in combination with multi-layer perceptrons, that exhibit remarkable adaptation in pattern classification applications. Various network topologies and sizes were tested in combination with different features’ sets. Quantitative analysis and validation results showed that the implemented system exhibits an overall recognition accuracy of 94.84%, while the error in separating bowel sounds from other sound patterns, representing interfering noises, was 2.19%.