Blood Cell Subtype Classification Using Deep Learning

Abstract

A vital component of human bodies is blood. Despite making up only 1% of the human body, blood cells have a greater impact on disease and infection detection. These blood cell subtype classifications also aid in ensuring that patients receive the quickest-acting medical treatment and recover swiftly. To get the findings, many machine learning techniques are used, including the Feature Extraction methods (K* classifier and Decision table) and the Image Segmentation algorithms (Random Forest and MLR). Training data sets are only a certain size since labelling data requires time and money, whereas unlabelled photos play a vital role in electronic medical record systems. The convolution machine learning process is replaced by the deep learning technique, which uses hidden data from unlabelled photos to sort them into different categories. The unlabelled image is subjected to data pre-processing, four-layered classification using the CNN algorithm (convolution layer, ReLu, pooling, fully connected layer), and multi-classification of blood types. The proposed work is found to categorise the white blood cell with 95% accuracy after thorough experimental investigation.