Using medical objects for clinical records classification

Abstract

In this paper, medical objects are used as features to classify clinical records.  Medical objects such as disease names, drug names, symptoms, examination  indicators are extracted using an Unstructured Information Management  Architecture (UIMA) based system.  The extracted medicalobjects will be used against the "bag-of-words" as the features of the clinical record in some classification algorithms. The results show that the precision of the classification results using medical objects is better in all algorithms, suggesting that medical  objects contribute a significant part to the semantic of a clinical record.