Using Aspect Ratio to Classify Red Blood Images

Abstract

In automated medical diagnosis, shape plays a key role in image processing and pattern matching. In particular, microscopic visual examination, as used in this paper, extensively uses shape to diagnose anemia using the shapes of Red Blood Cells (RBCs). This automated process depends entirely on the ability of the mammalian RBCs to change shape causing some types of anemia, which makes use of cell-discrimination on these varying RBC shapes an effective method of diagnosing anemia. In this automated diagnosis, image processing and pattern matching follows four steps: shape extraction, shape representation, shape size normalization and application of Fourier descriptors to obtain shapes of normal and abnormal RBCs. In testing the results, a client-server computer program applies invariant-moments to filter irrelevant shapes from the query, and aspect ratio of RBCs to improve on shape retrieval and verify results whereas increasing the geometrical number to reduces this accuracy by 10% to a promising 90%.