Auscultating Diagnosis Support System by Using Self-Organizing Map: Analysis of Long-Term Recording Medical Body Sounds

Abstract

Auscultation by a medical doctor means the diagnosis is made in a short time. Therefore, the unrecorded data and skill-dependent approach, changing an assessment over time through long-term diagnosis, is rather difficult. Thus, development of health monitoring equipment with long-term recording is desired. Previously, such a system was based on a self-organizing map (SOM) and a maximum entropy method (MEM) for data analysis, as suggested by the authors. SOM is an artificial neural network, which is trained using unsupervised learning in order to produce a feature map useful for visualizing the analogous relationships between input data. The authors recorded body sounds in order to produce a SOM. In this paper, we attempted the recording of body sounds over a long term period and assessed the changes over time. The result is that a SOM output is almost near areas (nodes)