White Blood Cell Classification Using Deep Transfer Learning

The count and the classification of WBC in the human body represent the state of the Immune System of a person and indicate the potential risks one can face. If there is any drastic change in the count of WBC from its standard, it implies that the human body is being affected by some disease. The WBC Classification serves as a good guideline in the identification process of deadly diseases. But the manual classification process is very slow and efficiency is proportional to the skills of expert doctors. Hence Computer-Based Automatic algorithms for will play a prominent role in the classification of WBC. In this paper, we proposed a transfer learning-based model for the unmanned classification of White Blood Cells based on blood smear images. The proposed transfer learning model is developed using various deep learning models such as VGG-16, Efficient Net, ResNet-50, Inception, and Xception. We have used the dataset Blood Cell Images from Kaggle which consists of around 12,500 images with sub-types Neutrophils, Lymphocytes, Basophils, Eosinophils, and Monocytes. Comparative analysis of the performance of each transfer learning-based model is performed. The transfer learning model developed using VGG-19 has achieved the highest accuracy of 92% among all compared models.