Analysis of White Blood Cell Segmentation Techniques and Classification Using Deep Convolutional Neural Network for Leukemia Detection

Abstract

Leukemia is a blood cancer which results due to inundation of abnormal white blood cells in human. Digital pathology is a dynamic condition that empowers extraction of precise and detailed information of amount and condition of WBCs from patient's blood smear. Digital pathology methods along with machine learning is gaining importance speedily as a proven and essential technology for investigation of new features of leukemia patients histological information, providing the reduced laboratory expenses, improved operational efficiency, diagnosis. This paper highlights emerging methods to automatically diagnose leukemia from cytological-histological & morphological analyses. In this study, we have performed comparative analysis of white blood cells segmentation techniques and evaluated the performance of pre-trained deep CNN with multiclass models for Support Vector Machine (ECOC) as feature extractors towards classification of WBC to assist in improved screening of leukemia with classification accuracy of 93.94%.