Comparison of Texture and Shape Features Performance for Leukemia Cell Images using Support Vector Machine

Abstract

Leukemia or often called blood cancer is one type of cancer caused by excessive white blood cells. Excessive white blood cells will cause disruption of normal function of other blood cells. To find out leukemia, we can do a physical examination in the form of a blood sample or can also use a spinal cord biopsy. In general, doctors take blood samples to see and look for abnormalities of the white blood cell count. To reduce human error in diagnosing leukemia, the study created two systems that can classify leukemia using the Hu moment invariant (HMI) and Support Vector Machine (SVM) methods and the Grey Level Co-occurance Matrix (GLCM) and SVM methods. Classification systems are used to classify acute and normal leukemia image classes using 10-fold cross validation in the sharing of its image data. The best classification results are the GLCM -SVM system with an accuracy value of 99% and the HMI-SVM system produces an accuracy value of 90%.