Red blood cell image enhancement techniques for cells with overlapping condition

Abstract

This paper proposes an automatic algorithm consists of techniques that can separate
overlapping RBC for enhancing RBC image to improve counting precision. The proposed
algorithm is able to automatically count RBC of colour blood smear images based on the
quality or challenging conditions of the images such as poor illumination of blood smear and
most importantly overlapping RBC. The algorithm comprises of two RBC segmentation that
can be selected based on the image quality, circle mask technique and grayscale blood smear
image processing. Detail explanations including experimental results that show the effectiveness of the proposed techniques are described in this paper. The proposed algorithm has been successfully detect and separate agglomerate RBC with 96% accuracy. Algorithm
validation was verified with RBC distribution area and confusion matrix.

Keywords: red blood cell (RBC); complete blood count (CBC); automated segmentation;
overlap RBC.