Efficient Features for Effectively Detection of Leukemia Cells

Abstract

Leukemia the disease of blood forming cells is a cancer that usually begin in the bone marrow. It results in high numbers of abnormal blood cells usually affecting the leukocytes, or white blood cells of a body. Oncologists and researchers are still working on appropriate reasons behind the cause of leukemia and its early detection as well. The contemporary techniques to detect leukemia are usually time consuming, laborious and subject oriented. In this research we presented a novel technique to detect the leukemia at its early stage. Detection of leukemia through images is quick and cost effective as there is no need of advanced lab testing and experts with in-depth knowledge. To identify whether the disease is acute or chronic, algorithmic techniques depend on the affected white blood cells. In our work, color filter is used as a preprocessing step to detect the region of interest that is white blood cells of ALL-IDB dataset. Then the structural feature (wavelet and curvelet descriptor) is used to detect the important features. This feature vector is trained on KNN and SVM classifier to check the correctness rate of this algorithm. Our approach achieved the accuracy of 92.7% which proved better in comparison with existing techniques in this field of medical research.