INTELLIGENT DECISION SUPPORT SYSTEM FOR RADIONUCLIDE DIAGNOSTICS IN CARDIOLOGY

Abstract

The algorithm of information-extreme machine learning of decision support system for myocardial perfusion scintigraphy with multilevel system of nested control tolerances for diagnostic feature values is designed. It is proposed the use of quantitative features which characterize the brightness of the pixels of the polar map radiopharmaceuticals distribution, and contextual features that describe the presence of symptoms, addictions and chronic diseases. The computing aspect of the normalized modification of information criterion of machine learning, which is a functional of the accuracy characteristics of decision rules, is considered. The results of machine learning with swarm optimization of control tolerances at different numbers of their levels are analyzed. Obtained an unmistakable rules on training matrix decision rules

Keywords

• перфузія міокарда, інформаційний критерій, система контрольних допусків, машинне навчання, категоріальні ознаки, розпізнавання образів