Acute Lymphoblastic Leukemia (ALL) is a human blood cancer which causes most deaths among other types of blood cancer. It commonly occurs in children and teenagers. ALL is recognised when there is a presence of lymphoblast cell in human blood. Haematologists diagnose ALL using visual microscopic observation from a human periphery blood sample. The inspection takes a long time and is limited to the subjective experience of the haematologist, and therefore, causes differences in the diagnostic results among haematologists. In this study, the researchers proposed automatic ALL Identification system to solve the problem. The method used an image processing technique that analysed a human blood cells morphology. The system identified the differences between normal white blood cells and lymphoblast cells by considering the shape and the size of the nucleus and also the cytoplasm. There are 33 microscopic an image data of a human blood cell that were used in this research. Geometric features used in the study were an area, perimeter, eccentricity, equivDiameter, solidity, roundness and circularity of a human blood the nucleus and cytoplasm and also the ratio of area and the perimeter of the cells and the nuclei. Firstly, the nucleus and cytoplasm were extracted from the other component of a human blood cell using Gram-Schmidt as an orthogonalisation-based segmentation method and then using Otsu threshold. Support Vector Machine was then used to classify whether the white blood cells were lymphoblasts or normal cells. Lastly, ALL identification was evaluated by the parameter of sensitivity and misclassification rate metrics using k-fold cross-validation.
|Journal||IOP Conference Series: Materials Science and Engineering|
|Publication status||Published - 8 Jan 2019|
|Event||1st International Conference on Advanced Engineering and Technology, ICATECH 2018 - Surabaya, Indonesia|
Duration: 29 Sept 2018 → …