Detection and classification of leukocytes in blood smear images: State of the art and challenges

Manual analysis of microscopic blood smears by highly expert pathologists is labor-intensive, time-consuming, and is subject to inter-observer variations. Recent innovations in image processing and computer vision techniques have improvised digital pathology in terms of objectivity and reproducibility. Traditional computer vision-based methods of recognition of white blood cell (WBC) from a pathological blood smear image includes the process of detection, segmentation, and classification. This paper presents a review of state-of-the-art detection, segmentation, and classification techniques for white blood cell analysis. The goal of this work is to present an introduction to the field, provide enough information about the analysis methods developed so far, and to be an appropriate reference for the researchers looking forward in this field. The methods under review are classified into intensity and feature based. The crucial steps involved in these techniques, mathematical foresights, performance evaluation techniques, issues, and future directions are discussed.