White Blood Cell Classification Using Convolutional Neural Network


Nahzat S., Bozkurt F., Yağanoğlu M.

Journal of Scientific Technology and Engineering Research, cilt.3, sa.1, ss.32-41, 2022 (Hakemli Dergi)

Özet

White blood cells (WBCs) are a key element of the immune system and demonstrate resistance to a variety of illnesses, quantitative and qualitative examination of various kinds of white blood cells is critical. Counting and categorizing the types of WBCs can help doctors detect and treat different illnesses. As a result, one of the most important steps in analyzing and testing blood samples is counting and categorizing various types of WBCs. The main purpose of this study is to provide a CNN based model for processing of WBCs with the aim of classifying the type of these cells. Kaggle white blood cells images were used in this article, we built a CNN-based model for classifying white blood cell types and assessed the model's performance using several optimizers. We have seen that the RMSprop optimizer shows the best result in our proposed model. We have compared four pre-trained models such as MobileNetV2, DenseNet121, InceptionV3 and ResNet50 with our proposed model. Compared to four pre-trained CNN models, and other related studies, our proposed model with the lowest number of trainable parameters and training time shows the great results with 99.5% accuracy, 99% recall, 99% precision, and 99% F1 score.