An Automated Segmentation and Counting of Ki67 Cells in Meningioma Using K-Means Clustering Technique

Abstract

Meningioma is a type of primary brain tumours. The meningiomas account for about one-third of all primary brain tumours. Image segmentation plays an important role in image analysis, especially detecting the tumours or cancerous areas in medical images. The output images from the segmentation prominently affect the system in detecting the tumour cells. Currently, the pathologists use the ‘eye-balling’ estimation technique to count the Ki67 cells. This technique was known as a time-saving measure. However, it has poor reliability and accuracy in counting the Ki67 cells. This paper proposed an automatic Ki67 cell counting in meningioma by using k-means clustering approach. The k-means clustering was used to segment the Ki67 cells and then the cells were classified into positive and negative Ki67 cells. The proposed system has been tested on 12 histopathological meningioma images. The proposed system is compared to the manually segmented images that have been validated in prior by the pathologists. The results show that the proposed system was able to segment the Ki67 cells with an average accuracy of 95.29%. The sensitivity and specificity of the proposed system were also high with an average of 93.56% and 97.39%, respectively