Image Segmentation using Enhanced K-means Clustering with Hybrid Image Filtering for Acute Leukemia Blood Cells Microscopic Images
In biomedical application, image processing become an interesting area that is considered as important role to perform further diagnosis or other task. Segmenting images is consider one of the important steps in image processing stage in digital image processing due to its wide spread usage and appl...
Saved in:
| Summary: | In biomedical application, image processing become an interesting area that is considered as important role to perform further diagnosis or other task. Segmenting images is consider one of the important steps in image processing stage in digital image processing due to its wide spread usage and applications. Recently many researchers have performed many research in assisting the haematologists to segment the leukocytes region from microscopic image of the blood cells in the issue of detecting the leukaemia cells in the early of prognosis. During the post processing, image filtering can cause some discrepancies on the processed image which may lead to insignificant result. The main objective of this research is to develop a method that capable to detect and segment the blood cell automatically for 100 microscopic images of patients suffering from acute leukaemia. The data was collected from the Department of Haematology, Universiti Sains Islam Malaysia, in Malaysia. Three clustering methods heve been utilised to perform the segmentation, which are Standard K-Means (SKM),Fuzzy C-Means and an enhanced method of K-Means algorithm (EKM) that facilitate the mean value of the k-centroids during initialization stage. Finally, image filtering system as background subtraction have been utilized namely Automated Thresholding (AT), Seeded Region Growing (SRG) and Mean Shift (MS) algorithm and the performances are analysed and compared to remove the background scene. The integrated clustering techniques using the Enhanced K-Means clustering together with hybrid image filtering system (MS-AT) as background subtraction have produced tremendous output images without the background scene. Experimental results shows a promising results performance of segmenting the blast cell of 100 microscopic image data using the EKM clustering and hybrid filtering system with the highest score of 99%. Briefly, this research present the advanced computational techniques for processing of microscopic images of blood samples from patients suffering from leukaemia. |
|---|