The discovery of sleep apnea with unsupervised neural networks

This paper presents extended Self-Organizing Maps (SOMs) for the discovery of sleep apnea patterns. The main idea lies in an adequate combination of several approaches where SOMs are used far temporal processing. Extended SOMs are part of a recently developed method temporal knowledge conversion that introduces several abstraction levels and generates a linguistic description of the discovered patterns. Such a representation form enables an evaluation of the results, since an intelligible description of the discovered patterns is obtained. An evaluation of the discovered patterns and their linguistic descriptions was made using a questionnaire. This method was successfully applied to a problem in medicine.