We propose a 3D version of the Contextual Multi-scale Multi-level Network (3D CMM-Net) with deeper encoder depth for automated semantic segmentation of different brain tumors in the BraTS2021 challenge. The proposed network has the capability to extract and learn deeper features for the task of multi-class segmentation directly from 3D MRI data. The overall performance of the proposed network gave Dice scores of 0.7557, 0.8060, and 0.8351 for enhancing tumor, tumor core, and whole tumor, respectively on the local-test dataset.
CITATION STYLE
Choi, Y., Al-masni, M. A., & Kim, D. H. (2022). 3D CMM-Net with Deeper Encoder for Semantic Segmentation of Brain Tumors in BraTS2021 Challenge. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 12962 LNCS, pp. 333–343). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-08999-2_28
Mendeley helps you to discover research relevant for your work.