Performance analysis of CNN fusion based brain tumour detection using Chan-Vese and level set segmentation algorithms

Abstract

Early diagnosis of a brain tumour may increase life expectancy. Magnetic resonance imaging (MRI) accompanied by several segmentation algorithms is preferred as a reliable method for assessment. In this study, first noise removed by median filter and dimensionality of datasets reduced by using random projection transformation (RPT). Next, the pre-processed images are clustered by using K-means and fuzzy c-means (FCM). In the very next step, the clustered images multi-features are fused by different data fusion approaches, and then segment the exact tumour area by using the active contour models such as level set method (LSM) and Chan-Vese (C-V). The performance of clustered based segmentation and fusion-based segmentation in terms of various fusion metrics. The results of both clustered based and fusion-based methods revealed that the CNN fusion-based segmentation performs better than clustered- based segmentation to detect the tumour with low segmentation error and minimal loss of information.