Classification of White Blood Cell Abnormalities for Early Detection of Myeloproliferative Neoplasms Syndrome Using Backpropagation

Fitri, Zilvanhisna Emka and Imron, Arizal Mujibtamala Nanda (2021) Classification of White Blood Cell Abnormalities for Early Detection of Myeloproliferative Neoplasms Syndrome Using Backpropagation. In: Proceedings of the 1st International Conference on Electronics, Biomedical Engineering and Health Informatics. Springer Nature Singapore Ptc Ltd, Singapore, pp. 499-508. ISBN 978-981-33-6926-9

	Text (Hasil Similarity) 1. LNEE_Classification of White Blood Cell Abnormalities for Early Detection of Myeloproliferative Neoplasms Syndrome Using Backpropagation.pdf - Supplemental Material Available under License Creative Commons Attribution Share Alike. Download (1MB)
	Text (Full Paper) 1.LNEE_2021.pdf - Published Version Restricted to Repository staff only Available under License Creative Commons Attribution Share Alike. Download (2MB)

Abstract

One of the diseases of white blood cells is myeloproliferative neoplasms syndrome where this disease is an abnormality in the bone marrow in excessive blood cell counts. Full Blood Count (FBC) is a type of examination that shows the patient’s health status, blood cell abnormalities, and the presence of infection in the patient’s body.However, the determination of cell abnormalities is still done manually based on the knowledge and experience of clinical pathology so that the determination of these abnormalities is subjective. Therefore we need a system that is able to classify white blood cell abnormalities automatically, objectively and accurately. This study uses digital image processing on peripheral blood smear images then feature extraction will be obtained which will be the input of the classification system. The features used are area, perimeter, metric and compactness, while the classification method used is the backpropagation method. The best backpropagation network architecture used is 4, 6, 8, 5, 2 with a variation of learning rates of 0.05 and 0.3 producing the best accuracy rate of 91.82% with the amount of training data that is 516 and testing data is 159.

Actions (login required)

View Item