Blood Cell Images Segmentation using Deep Learning Semantic Segmentation

TLDR

Segmentation of red and white blood cells in peripheral smear images is crucial for diagnosing disorders, yet remains challenging despite common image‑processing enhancements. This study applies deep‑learning semantic segmentation to segment red and white blood cells in blood smear images. The authors employ a deep‑learning semantic segmentation architecture to segment RBCs and WBCs directly from raw smear images. The model achieved an overall accuracy of 89.45%, with class‑specific accuracies of 94.93 % for WBCs, 91.11 % for RBCs, and 87.32 % for background.

Abstract

Segmentation of red blood cells (RBCs) and white blood cells (WBCs) in peripheral blood smear images plays an important role in the evaluation and diagnosis a vast of disorders, including infection, leukemia, and some types of cancer. Generally, various image processing techniques are used to enhance the quality of images before the segmentation step. Therefore, the segmentation of blood cells is still a challenge. However, in this research, deep learning semantic segmentation - cutting-edge technology is applied for segmentation red blood cells and white blood cells in blood smear images. The experiment result shows that the global accuracy of our model yielded 89.45%. Besides, the accuracy of the segmentation of white blood cells, red blood cells, and the background of blood smear image reached 94.93%, 91.11%, and 87.32%, respectively.