Investigating size features of acute leukemia using k-nearest neighbors

Abstract

This study investigates the size feature of acute leukemia of blood sample image by using k-NN as a classifier. K-NN is believed to be one of the simplest methods in classifying data including images. Hence, this paper manipulates the size feature by varying the combination of sub features. From here, the capability of size feature in classifying acute leukemia blood sample images into AML and ALL are tested. Results show that, even though the highest accuracy is not achieved by the combination of all size sub features but it still could obtain up to 98.67% of accuracy. The best accuracy obtained is 99.78% by combining radius and perimeter as an input feature. Therefore, k-NN could be applied as a classifier for this classification problem.