Artificial Intelligence in Digital Holographic Imaging

14Automated Phenotypic Classification of Red Blood Cells

14.1 Introduction

Human blood contains different cell types. RBCs or erythrocytes are the most abundant cell type. They transport oxygen to the tissues and organs as well as carbon dioxide to be removed by the lungs. The biconcave shape of the erythrocyte is extremely important for RBC functionality as it increases the surface‐area‐to‐volume (SAV) ratio and facilitates the large reversible elastic deformation of the RBC required to squeeze through tiny capillaries [1]. Pathological disorders can modify RBCs and lead to significant changes in their original shape [2]. The consequences of modified RBCs are often observed as clinical symptoms that range from the obstruction of capillaries and restriction of blood flow to necrosis and organ damage [2, 3]. Counting cell types in a blood sample during cytometry is an important task to investigate clinical status.

In the case of RBCs, a biconcave cell type accounts for a substantial portion of RBCs in a healthy person, although there are other RBC shapes with different percentages between healthy and non‐healthy persons [4]. Accordingly, it is essential to determine the percentage of each RBC type in a blood sample that contains different RBC shapes to diagnose and determine the appropriate treatment of subjects. Typically, an image‐based cell analysis for diagnosis is performed by experts. It has drawbacks, like being time‐consuming and inaccurate. A sample is generally viewed ...