White Blood Cell Classification Using Conventional and Deep Learning Techniques
P. Pandiyan, Rajasekaran Thangaraj, S. Srinivasulu Raju, Vishnu Kumar Kaliappan, B. Lalitha · 2022
White blood cells (WBCs) play an essential part in examining the health status of an individual. The decision about blood disease diagnosis requires recognition, detection, and characteristics of a blood sample under test. Manual assessment methods-based blood cell classification leads to intra- and inter-level changes in the results acquired by haematologists. As a result, the automatic classification of WBCs is of great importance in the area of medical image analysis. Currently, automatic classification of blood cells is done through blood smear images. The conventional WBC classification system follows a group of tasks to be performed sequentially. This motivated us to adopt deep learning to classify blood cells with raw images and enhance prediction accuracy. The blood smear gives vital information on disease when it is visualized under a microscope. This analysis is instrumental in diagnosing many diseases, such as malaria, leukaemia, anaemia, etc. The WBC count decides the disorders, and structural differences like size, colour, and shape variations aid in disease diagnosis. Assessment of WBCs through manual means provides data such as morphological analysis and total and differential WBC count. The disadvantages of this manual evaluation have led to observational errors and laborious tasks. Computer-based WBC evaluation automatically reduces haematologists’ workload. Blood cell classification through automatic methods provides fast results and also handles massive data effectively. Automated classification of WBCs can be done in two ways, namely conventional and deep learning methods. In this chapter, a review of the classification of blood cells using traditional and deep learning methods is discussed.