A Classification of Platelets in Peripheral Blood Smear Image as an Early Detection of Myeloproliferative Syndrome Using Gray Level Co-Occurence Matrix

Abstract

Platelet disease is usually caused by abnormalities of the number or form of platelets, for example in Essential thrombocythemia (one of the groups of myeloproliferative syndrome). The characteristic of ET disease is that if a lot of giant platelets are found as large as leukocytes and cannot be detected using FBC, microscopic examination must be done manually by a clinical pathologist. The classification process begins with image processing techniques on the peripheral blood smear image, then texture features are taken using the Gray Level Co-Occurrence Matrix ( GLCM) which consists of ASM, IDM and Entropy features.This feature is input into the classification system using Backpropagation. The test results, Backpropagation was able to accurately identify cells in BG images, namely leukocytes 91.84%, normal platelet cells 92.86% and giant platelet cells 84.69%. Whereas in the AL image, the accuracy of leukocyte cells is 90.82%, normal platelet cells are 96.94% and giant platelet cells are 87.76%. The average accuracy of the Backpropagation method at 84.69% BG images and AL images was 87.76%. So this classification system is able to be used as a tool for doctors or medical analysts to speed up the process of early detection, especially in myeloproliferative syndrome patients.