Comparative Analysis of Thresholding Methods in Cancer Cells Image Processing

Abstract

Analysis of cancer cell images in medical field needs to be assisted using digital image processing. This paper presents the comparative analysis of image thresholding using two algorithms, Backtracking Search Algorithm and Particle Swarm Optimization. Two experimental designs were implemented. In the first design the images were thresholded and the performance was compared. In the second design, the images were enhanced before the thresholding was performed. In the second, the original and processed image histograms were presented and compared. In both designs, performance metrics were calculated to validate the comparative analysis. In the first experimental design, where BSA and PSO are implemented to threshold lung epidermoid carcinoma and chronic lymphocytic leukaemia cell images, the values of MSE, PSNR, MSSIM, FSIM and IEM show the superiority of BSA over PSO. In the second one where the thresholding method is implemented after image enhancement process, the histogram entropy and variance show that the thresholding method using BSA outperforms the one using PSO. These results show that in both designs the BSA outperforms PSO. Therefore, the thresholding method using BSA is more suitable for cancer cells image thresholding in processing the image samples for further analysis. This will provide a more reliable solution and effective way for assisting analysis of cancer cells where it minimize the difficulties arises in the conventional way of manual observation of microscopic images.