Dual modality search and retrieval technique analysis for leukemic information system.

Abstract

Medical Information System (MIS) deals with standardized method of collection, storage, retrieval and evaluation of patient data. Computational Intelligence (CI) technique has many abilities in data processing and structuring, pattern matching, forming knowledge base, reasoning and decision making. MIS for leukemia cells was developed at University of Malaya for the Hematology Department of Hospital Ampang, Malaysia. CI techniques were used for syntactical and contextual image retrieval of leukemic images. A logical approach was used to systematically develop content based image retrieval system that supports decision making in clinical hematopathology. The developed system is based on open source. Pages were designed and developed using Java Server Pages (JSP) and Java with MySQL as the database for the domain and image repository. Several Java-based tools were used for image processing, neural network based pattern classification and recognition, and for the hybrid system. The classification rates for individual white blood cells varied between 97% for training data and 95% for testing data depending on the individual cell type- monocyte, lymphocyte, eosinophil, basophil and neutrophil. This result is cross-validated by an expert hematologist. Key words: Medical information system (mis), leukemia image retrieval, image processing, feature classification, pattern recognition, hematology.