Automatic Early Detection and Classification of Leukemia from Microscopic Blood Image

Abstract

Leukemia is a form of blood cancer that affects white blood cells, and is one of the leading causes of death among humans. Currently, diagnosis of leukemia is done through visual inspection of microscopic images of blood cell, which is time consuming, tedious, and requires trained human experts. Therefore, the lack of an automatic, early, and effective leukemia detection system is a great challenge in Ethiopian hospitals. The main objective of this research is to develop an automatic early detection, and classification system to diagnose leukemia from blood image using machine learning and image processing algorithm. To do the research, 400 leukemic blood images and 50 normal blood images had acquired from Jimma University Specialized Hospital using digital microscope, and preprocessed with contrast enhancement. K-means image segmentation and feature extraction were applied by the system. Multi Class Support Vector Machine has used to provide detection and classification of leukemia disease based on the extracted features parameter. The leukemia disease detection and classification accuracy achieved by developed system is 94.62%. Moreover, 94.17% sensitivity and 100% specificity level has been gained by the system. It takes an average of one minute to provide the diagnosis result. The potential of digital image analysis for leukemia disease diagnosis using artificial intelligence; which is not tedious and time consuming is very beneficial when compared to the manual method. In the future, direct diagnosing system of leukemia without staining process is recommended.