A Deep Learning-based Approach for WBC Classification
K S Ramyashree, B. Sharada, R. Bhairava · 2024
Human immune systems rely on white blood cells (WBCs) to fight against disease-causing pathogens. Blood cells contain vital information that can be used to determine a person’s health status. An efficient and accurate classification of blood cell types is critical due to their different features. In the domain of histopathology, WBCs hold a significant role. The analysis of these WBCs can provide valuable insights to healthcare professionals, aiding in diagnosing conditions such as tumors and infections caused by viruses. In this work, EfficientNetB0 is proposed as an advanced deep learning (DL) framework built on artificial intelligence (AI) designed for the WBC’s automatic categorization, which contains five distinct classes. These classes include basophils, eosinophils, neutrophils, lymphocytes, and monocytes. This work demonstrates the most effective efficiency, achieving WBC-type classification in significantly fewer epochs and less time than alternative methods. The approach that has been suggested is assessed using the dataset, which is available on the public Raabin Health Database, to evaluate its performance. The best performance in the assessment has been achieved through applying EfficientNetB0, with 99.17% overall efficiency and a 99.0% F1-score. In contrast to the latest DL models, the EfficientNetB0 model is designed to achieve excellent outcomes.