Automatic brain tumor segmentation is an ill-posed problem. In this paper, we propose convolution neural network based approach for brain tumor segmentation. The proposed network made up of encoder-decoder modules. The encoder modules are designed to encode the input brain MRI slice into set of features while the decoder modules for the generation of the brain tumor segmentation map from the encoded features. To maintain the structural consistency, feature maps obtained on the encoder side are shared with the respective decoder modules using skip connections. We have used training set of the BraTS-15 dataset to train the proposed network for brain tumor segmentation. While, its testing set is used to validate the proposed network for brain tumor segmentation. The experimental analysis consists the comparison of the proposed and existing methods for brain tumor segmentation with the help of Dice similarity coefficient and Jaccard index. Comparison with the existing methods show that the proposed method outperforms other existing methods for brain tumor segmentation.
CITATION STYLE
Bhalerao, K., Patil, S., & Bhosale, S. (2022). Brain Tumour Segmentation Using Convolution Neural Network. In Communications in Computer and Information Science (Vol. 1567 CCIS, pp. 187–197). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-11346-8_17
Mendeley helps you to discover research relevant for your work.