A case-based approach for classification of physiological time-series

Abstract: Complex measurement classification is often difficult, as in the medical domain, and it usually takes a long time to fully master all aspects involved. An automated measurement classification system would ease the diagnostic process for treatment personnel, especially for less experienced clinicians. This thesis contains results from research in the field of Artificial Intelligence (AI) applied to medical measurement classifications. Artificial Intelligence may be described as a variety of computational methods and techniques that exhibit intelligent behaviour. These methods and techniques enable problem solving comparable to humans. The thesis presents a novel approach for multiple time-series analysis based on Case-Based Reasoning (CBR). CBR is an AI method based on a plausible cognitive model of human reasoning. The approach analyses parallel streams of measurements and uses CBR as well as other AI methods for classification and domain reduction. The approach is implemented as a system for classification of Respiratory Sinus Arrhythmia (RSA). The time-series are composed of physiological measurements as the system identifies dysfunctions within the RSA. RSA is identified by analysing the heart and the pulmonary systems of the human body. The developed system, named HR3modul, functions as a decision support tool for treatment personnel in the field of psychophysiological medicine. A classification proposal is presented to the user. The proposal is based on stored knowledge and current physiological time-series.