Application of Deep Learning in Counting WBCs, RBCs, and Blood Platelets Using Faster Region-Based Convolutional Neural Network
Nirav Jain, Shail Shah, Ramchandra Sharad Mangrulkar, Pankaj Sonawane · 2021
Convolutional neural network (CNN) is a part of deep neural network, mainly aimed at developing machine learning (ML) models on visual imagery, which is also referred to as the shift-invariant artificial neural network due to a shared weight architecture. One such application that this chapter will be discussing is the CBC test using faster region-based convolutional neural network (RCNN). The blood consists of three major types of cells: RBCs or erythrocytes, WBCs or leukocytes, and blood platelets or thrombocytes. Traditionally, techniques such as “Coulter Counters and Laser Flow Cytometry” were used, which required a very complicated, costly, and time-consuming system for performing the CBC test. The work presented uses the power of a CNN to develop the model to count the blood cells. It uses an optimized version of CNN, known as the Faster RCNN, which processes the images quickly and efficiently. Convolutional function is a linear function used as an activation function in CNN.