View article

White blood cells play important rule in the human body immunity and any change in their count may cause serious diseases. In this study, a system is introduced for white blood cells localization and classification. The dataset used in this study is formed by two components, the first is the annotation dataset that will be used in the localization (364 images), and the second is labeled classes that will be used in the classification (12,444 images). For the localization, two approaches will be discussed, a classical approach and a deep learning based approach. For the classification, 5 different deep learning architectures will be discussed, 3 pretrained architectures and 2 customized architectures will be presented. After discussing this models and test them on the dataset, the best selected model will be evaluated describing the obtained results. The localization module achieved average Intersection over Union (IoU) of 71%, while the classification module achieved 92% classification accuracy. In addition to reporting the model performance, the model robustness was also checked by adding three different types of noise, Gaussian noise, salt and pepper noise, and speckle noise. This system outperforms other studies in the literature, where the accuracy was either less than the obtained from the system or the dataset was much smaller the used data in this study.