Abstract—Blood cells are one of the most important parts in humans. One type of blood cells that play an important role in a leukemia diagnosis is leukocyte cells. There are some types of leukocyte i.e. myeloblast, lymphoblast, monoblasts and erythroblasts. One method of measuring leukocyte cell abnormalities is by examination of the morphology of leukocyte cells covering the area, circumference and diameter of leukocyte cells. In this research will be identified morphology of myeloblast cells by using K-means clustering method. The observed control variable is a characteristic of myeloblast cell that includes diameter, contour, and uniformity of object, amount, and cell density. While the observed data is uncontrolled image (noise) with RGB color format. The experiment showed promising results for further development. Index Terms—Leukocyte cell, leukemia, morphology, K-means clustering, uncontrolled image. I. INTRODUCTION Blood is a fluid contained in the body of living things except plants, which serves to deliver substances and oxygen to the tissues of the body [1]. One type of blood cell is a leukocyte cell or commonly known as leukocytes. Abnormalities in leukocytes can refer to one type of fatal disease that is blood cancer or commonly known as leukemia. The diagnosis of leukemia is based on clinical signs and symptoms and investigations. Clinical signs and symptoms commonly found are malaise, fatigue, fatigue, pallor, frequent infection, heat, joint pain, full race in the abdomen, or abnormal bleeding. Laboratory investigations used to assist in the diagnosis are complete blood tests, peripheral blood morphology, bone marrow morphology, cytogenetics, DNA, and cluster of differentiation detection using flow cytometry [2]. For the developing countries such as Indonesia, Most laboratories still use cell morphology to help diagnose leukemia due to limited resources, both infrastructure and human resources. This check is cheaper. However, morphological examination requires the expertise of a limited number of clinical pathologists. This examination is sometimes less valid because in some cases it is difficult to distinguish blast cell morphology into the type of myeloblast, lymphoblast, monoblasts, or erythroblasts so that potential errors of diagnosis. Under the above conditions, it is Manuscript received October 19, 2018; revised April 17, 2019. Retno Supriyanti, Ahmad Rifai, and Yogi Ramadhani are with Electrical Engineering Dept, Jenderal Soedirman University. Kampus Blater, Jl. Mayjend Sungkono KM 5, Blater, Purbalingga, Jawa Tengah, Indonesia (e-mail : retno_supriyanti@unsoed.ac.id). Wahyu Siswandari is with Medical Department, Jenderal Soedirman University, Jl. Gumbreg, Purwokerto, Jawa Tengah, Indonesia. necessary to develop an automatic blood cell type detection device as a low-cost, easy-to-use and accurate leukemia diagnostic tool so that it can be distributed across all healthcare units throughout Indonesia and in particular to remote areas. One solution to solve this problem is the use of digital image processing techniques for morphology identification of leukocyte cells. In our previous research [3]-[9] we developed simple and easy-to-use system for handling problems about health services in rural areas by implementing image processing techniques in the case of cataract diseases, pregnancy diagnosis, and dental segmentation. There are several studies that use digital image processing techniques in the identification of blood cells. Putzu [10] in his research conducted segmentation using thresholding. One post segmentation process is the grouping of single objects and pitch object as well as the separation of coincident cells. Ajala [11] conducted a comparative study for edge-detection and watershed-based segmentation analysis. Edge detection methods are used to obtain ridges, lines and contours along red blood cells. While the Watershed method includes opening and closing reconstruction on the overlapping images. The results show that segmentation by using watershed method is better than edge-based detection. Higgins [12] analyzed the relationship of blood flow and the environment due to varied blood cell variations. He uses image morphology computations and machine learning algorithms to measure fluctuations in flow velocity. Stadelmann [13] identifies an automatic calculation number of leukocyte cells using the Ada Boost method. The result is a real time automatic system to calculate the number of leukocyte cells in a scan using a microscope. Guo [14] did research with leukocyte cell in bone marrow using multispectral imaging technique. For image segmentation he uses the Support Vector Machine (SVM) which is applied directly to the spectrum of each pixel of the microscopic image. Gual-Arnau [15] did research on the red blood cell object in particular is a sickle cell. These sickle cells cause erythrocyte-containing hemoglobin polymerization. He focused his research on sickle cell shape changes using integral-geometry method, active-contour segmentation method and k-NN classification method. Alferez [16] developed a method of introducing various types of lymphoid cells automatically. In his research he used the component clustering and watershed transformation in segmenting the image. Kaewkamnerd [17] in his research, he developed automated equipment for the detection and classification of malarial parasite species using the image of blood cells. The system he developed uses image-processing analysis combined with motorized unit hardware mounted on a microscope. This is done to get a quality microscopic image Characteristics Identification of Myeloblast Cell Using K-Means Clustering for Uncontrolled Images Retno Supriyanti, Ahmad Rifai, Yogi Ramadhani, and Wahyu Siswandari International Journal of Machine Learning and Computing, Vol. 9, No. 3, June 2019 351 doi: 10.18178/ijmlc.2019.9.3.809