Convolutional Neural Networks Learning from Respiratory data

Convolutional neural networks (CNNs) have been successfully applied in a wide variety of fields, from image processing to genomic sequencing. In the context of biomedical data, we focus our attention on respiratory data, complex signals characterized by a high level of information richness and potential indicators of several common medical conditions. In this respect, we address the problem of identifying unhealthy indicators in respiratory sound data through the application of novel CNNs architecture and the extraction of Mel Frequency Cepstral Coefficients (MFCC), in order to unveil unhealthy indicators and provide doctors with a potentially life-saving tool.