Brain tumor segmentation is an essential and challenging task because of the heterogeneous nature of neoplastic tissue in spatial and imaging techniques. Manual segmentation of the tumor in MRI images is prone to error and time-consuming tasks. An efficient segmentation mechanism is vital to the accurate classification and segmentation of tumorous cells. This study presents an efficient hierarchical clustering-based dense CNN approach for accurately classifying and segmenting the brain tumor cells in MRI images. The research focuses on improving the efficiency of the segmentation algorithms by considering the qualitative measures such as the dice score coefficient using quantitative parameters such as mean square error and peak signal to noise ratio. The experimental analysis states the efficacy and prominence of the proposed technique compared to other models are tabulated within the paper.
Magnetic resonance Imaging (MRI) is one of the most utilized medical imaging techniques for detecting and diagnosing the different abnormalities such as tumors, lesions within various internal organs of human body. Manual segmentation from these images is timeconsuming and complex. Due to this, automatic segmentation techniques such as convolutional neural network (CNN), deep neural networks are used. An accurate, efficient and advanced computational segmentation method is extremely required for fatal diseases like brain tumor from brain MRI images. In this research work, deep learning model comprising of three channels such as patch extraction, 6 layered transfer-learning capsule and 5 layered segmentations capsule This proposed work addresses deep learning coupled with small kernels and handles the obstacles in brain tumor segmentation techniques. The proposed model is a 11-layered deep capsule network consisting of transfer learning along with two dropout layer at the input and 5 layers of segmentation channel along with dropout layers. The work presented in this paper focusses in attaining a high dice score coefficient and accuracy in brain tumor segmentation from MRI images. The model presented in this paper is effectively trained over 150 images in the dataset. The proposed work has attained comparative better results with respect to the dice score coefficient such as (0.80, 0.76, 0.76) for whole tumor [WT], active tumor [AT] and core tumor [CT] respectively.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.