An Automated Segmentation and Counting of Ki67 Cells in Meningioma Using K-Means Clustering Technique
View/ Open
Date
2019Author
Fahmi Akmal, Dzulkifli
Mohd Yusoff, Mashor
Hasnan, Jaafar
Metadata
Show full item recordAbstract
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