Facilitating Leucocyte Count Using Deep Learning: A Paradigm Shift

White blood cell classification is a task that is given paramount importance in the field of pathology in order to accurately diagnose a plethora of ailments and diseases. Blood cell classification is essential for the calculation of the differential count of blood cells in a particular blood sample. Physicians derive a medical prognosis based on blood reports in which differential count plays a key role. To reduce the cost and difficulty of this task as well as to improve accuracy in cell classification, image processing and deep learning have proven to be viable alternatives to conventional pathological methods. To tackle this problem through image classification, we have successfully conducted a comparative study of accuracy metrics of various Convolution Neural Network models including the very popular Alexnet, VGG, ResNet, Inception, Densenet and EfficientNet. With multiple methods of preprocessing the dataset and deploying these well-known image classification models, the most accurate and computationally inexpensive deep learning model was identified for WBC classification.