Robust Cell Detection for Automatic Ki-67 Counting Using Adaptive Fuzzy C- Means Clustering and Classification

Abstract: Neuro endocrine tumor (NET) is one of the most common cancers leading to death worldwide. Many studies have emphasized the importance of Ki-67 as the proliferation marker in the neuro endocrine tumor. Automatic Ki-67 assessment is very challenging due to complex variations of cell characteristics. In this paper, we propose an integrated learning based framework using Fuzzy C means clustering (FCM) for accurate automatic Ki-67 counting for NET and to localize both tumor and non-tumor cells. Unlike the non-fuzzy clustering algorithms, FCM is less sensitive to noise and give better results for overlapped data sets. For feature selection t-test algorithm is used. The t-test has been used to rank features for microarray data. For multi-class problems, t - statistics value for each gene of each class is calculated by evaluating the difference between the mean of all the classes, where the difference is standardized by the within class standard deviation. The automatic Ki-67 counting is quite accurate compared with pathologists' manual annotations. This is much more accurate than existing methods. Keywords: Neuro endocrine tumor, cell detection, classification, Ki-67, segmentation.