Computer-aided screening for Acute Leukemia blood infection using gray-level intensity

Screening for Acute Leukemia is critical task just like other health diagnosis. Some computerized methods which do not need domain experts' involvement in decision making have been proposed. These methods main challenge lays in the image processing part of system design to isolate the lymphocyte from other blood cells while taking care of various morphological shapes of the lymphocytes. In this paper we propose a method to segment the lymphocyte from other white blood cells; the proposed method can take care of various shapes of lymphocytes. We use the segmented lymphocyte to extracted gray intensity features for building a classifier. The proposed method takes microscopic image as input data for processing and outputs diagnosis results saying that the patient's blood image was saint or has leukemia. The contribution brought by our proposed method is the ability to segment lymphocyte image from noisy image while maintaining both nucleoli and cytoplasm parts intact for the classification to be possible using only gray-scale intensity distribution. The experimentation resulted in 93.7% correctly classified, mean error 0.08, false positive 0.07 using Decision tree with cross-validation folds 10.