Brain tumour segmentation in flair mri using sliding window texture feature extraction followed by fuzzy c-means clustering

Abstract

In this paper, a hybrid approach using sliding window mechanism followed by fuzzy c means clustering is proposed for the automated brain tumour extraction. The proposed method consists three phases. The first phase is used for detecting the tumorous brain MR scans by implementing preprocessing techniques followed by texture features extraction and classification. Further, this phase also compares the performance of different classifiers. The second phase consists of the localization of the tumorous region using sliding window mechanism, in which a sized window sweeps through the whole tumorous MR scan and the window is classified as tumorous or non-tumorous. The third phase consists of fuzzy c means clustering to get the exact location of the tumour by removing the misclassified windows obtained from Phase 2. 2D single-spectral anatomical FLAIR MRI scans are considered for experiment. Outcomes demonstrate significant results in terms of sensitivity, specificity, accuracy, dice similarity coefficient in comparison with the other existing methods.