Classification of blood cells into white blood cells and red blood cells from blood smear images using machine learning techniques

Classification of blood cells from Peripheral Blood Smear (PBS) images is a crucial step to diagnose blood-related disorders such as leukemia, anemia, an infection, cancer, and polycythemia. In blood cell-based analysis, the hematologists always make a decision based on the total number of cells, their morphology, and distribution using a microscope. Hematology analyzer, flow cytometry provide reliable and exact Complete Blood Count (CBC) indicating abnormalities in the blood smear slide. The methods being used are very expensive, timeconsuming, require manual intervention and not accessible in many medical centers. Therefore there is a necessity for an automatic, inexpensive and robust technique to detect various types of diseases from any PBS images. The automatic classification model improves the hematological procedures, quickens the diagnosis process and enhances the accuracy of the evaluation process. Thus in this paper, we used a semi-automatic method to segment and classify blood cells into White Blood cell (WBC) and Red Blood Cell (RBC). Texture features of a cell are extracted using Gray Level Co-occurrence Matrix (GLCM) and fed to the classifiers like Naive Bayes classifier, K-nearest neighbors, decision tree, K-means clustering, random forest, logistic regression, ANN and SVM. The performance parameters are compared and found that the logistic regression is best suited for the work with the 97% accuracy.