Automatic cell nucleus segmentation using superpixels and clustering methods in histopathological images

It is observed that cancer and cancer-related deaths increase day by day. Early diagnosis is vital for the early treatment of the cancerous area. Unhealthy cells, which are diagnosed by expert pathologists as a result of efforts, provide early detection with the help of computer-aided programsIn this study, kMeans and Fuzzy C Means methods, which are among the global segmentation methods, and SLIC, Quickshift, Felzenszwalb, Watershed and ERS algorithms, which are among the superpixel segmentation methods, were used for automatic cell nucleus detection in high resolution histopathological images with computer aided programs. It is seen that better success is obtained in kMeans and FCM algorithms in high resolution histopathological images used as a result of the study. Quickshift and SLIC methods gave better results in terms of precision. In the F-Measure (F-M), it is seen that the best success is the k Means and FCM algorithms and the true negative ratio (TNR) is more successful in Quickshift and SLIC methods.