Separable CNN based Automated Classification of WBCs in Peripheral Blood Stream

The proportional percentage of each kind of white blood cell (WBC) in a blood sample is determined by blood cell classification, which is a popular diagnostic test. This test, on the other hand, is carried out by pathologists visually inspecting a blood sample, which is a time-consuming and difficult procedure. This test can also be done automatically with the right equipment. However, such equipment is costly and only available at major medical facilities. In this study, an alternate strategy for WBC detection and identification in a blood picture is provided, which is based on a low-cost microscope and digital camera combined with a separable convolutional neural network. The suggested method promises to generate 99.71 percent accuracy and 98.64 percent precision, respectively. It not only utilizes erratic optimizers and learning rate but also uses a large database of more than 10,000 samples.