Deep transfer learning-based classification of White Blood Cells using customized classification base

White blood cells (WBCs) are essential for our immune system to fight against infections and foreign agents. However, previous research has limitations in accurately classifying different types of WBCs, which can indicate severe diseases like leukemia. On the other side Transfer learning (TL) has become a popular research choice for medical image analysis due to its promising results. The pre-trained state-of-the-art Convolutional Neural Networks (CNNs) have produced highly competitive results compared to other techniques. This paper highlights the classification performance of the deep transfer learning-based approaches, when applied to a well-known publicly available white blood cell (WBC) dataset, the Rabbin-WBC dataset. The paper aim is to investigate the performance of the proposed customized FCNN while used with the vision of transfer learning and with and without tuning of pre-trained models. The results, which we are confident in, reveal that the customization done in the classification part of the classical architectures, ResNet50, InceptionV3 and InceptionResNetv2, for the classification of WBCs achieves accuracy of 0.98.