White Blood Cells Identification and Classification from Microscopic Blood Images

Abstract

Identification of issues related to blood can be done by the visible assessment of components microscopic images present in the blood. From the recognition of blood issue, it can prompt the identification and classification of many diseases relevant to blood. Analalysis of white blood cells of allows for the detection of Acute Lymphoblastic leukemia, it is a type of blood cancer that can result in death if not treated and detected early. In this work, a system is proposed for identification and classification of white blood cells using microsccopic images of blood.
This process can also be performed manually by skilled operators but it have various drawbacks such as slow process, lengthy calculations, non accurate results high computational cost. Various system are already existed but some of them are partially developed.
The proposed work consists of various phases that uses microscopic images of blood taken from database ALL IDB. This database consists of microscopic blood images of humans and patients suffering from leukemia. This work starts with WBCs detection and segmentation of adjacent WBCs after this feature extraction is done, here shape and texture features are extracted that is used to train the classifier to determine which classifier is best for classification of leukemia.
Result obtained from SVM classifiers are compared with other classifier such as K-NN and BPNN also SVM is tested with common kernels and we can conclude from the results that SVM gives better result with accuracy 96.97 % for GLRLM features.