Diagnosis of Acute Lymphoblastic Leukemia Using Microscopic Blood Cell Images

Abstract

Abstract: Acute Lymphoblastic Leukemia (ALL) is a kind of blood illness attributable to the surprising ascent in the development of unhealthy WBCs in the spongy tissues of the bone, prompting blood malignant growth. It may be seen in children and aged people too.

    The study of microscopic images proves a substantial role in the evaluation of leukemia and its effective detection.  The existing techniques are of traditional type and somewhat depend on human intervention, which is laborious. So, an automated leukemia diagnostic system is very much needed that reduces manual intrusion and provides more precise medical information.

   This paper describes an automated system developed, based on image processing techniques for the detection of acute lymphoblastic leukemia in blood cells. Here a system is proposed to detect ALL by examining microscopic blood cell images obtained from a standard dataset. In our research work, two image processing techniques are suggested for the detection of the illness. The first technique depends on the conventional feature extraction procedure where the features like region, edge, quantity of nuclei etc., are separated. Data is then sent to the classifier to be categorized. Prior to feature filtration, the images are processed by an adaptive k-means segmentation algorithm to separate the nucleus. The input image is fed to DNN in the other technique.

 The overall appraisal of the presentation of classifiers like SVM ANN is performed with features obtained from the first technique. The first technique furnishes a detection efficacy of 89.37% with SVM and 92.16% with ANN. The CNN-dependent feature extraction technique offers a detection efficacy of 93.36%.

Index Terms: Microscopic Blood Cell Images, acute Lymphoblastic Leukemia (ALL), SVM, ANN, CNN, automated leukemia detection