Establishing Medical Diagnosis using Pattern Semantic Rules

Abstract

The developed methods are based on pattern rules to support medical image diagnosis. They have important characteristics that make them different from other CAD methods: the process is completely automatic, with the possibility to define a great number of diagnosis; the methods could be applied to any medical domain, because the visual features, the semantic indicators remain unchangeable, and the semantic rules are generated by learning from labeled images-examples; the selection of the visual characteristics set is based on their retrieval accuracy; the spatial information of the regions is considered, offering important medical information as the relationships of a sick region with another. Although we present the results achieved in endoscopic images analysis, our methods can be used to analyze other types of medical images. The prototype system was applied to real datasets and the results show high accuracy. Ill. 2, bibl. 7 (in English; summaries in English, Russian and Lithuanian).