Automated segmentation in brain magnetic resonance image (MRI) plays an important role in the analysis of many diseases and conditions. In this paper, we present a new architecture to perform MR image brain segmentation (MRI) into a number of classes based on type of tissue. Recent work has shown that convolutional neural networks (DenseNet) can be substantially more accurate with less number of parameters if each layer in the network is connected with every other layer in a feed forward fashion. We embrace this idea and generate new architecture that can assign each pixel/voxel in an MR image of the brain to its corresponding anatomical region. To benchmark our model, we used the dataset provided by the IBSR 2(Internet Brain Segmentation Repository), which consists of 18 manually segmented MR images of the brain. To our knowledge, our approach is the first to use DenseNet to perform anatomical segmentation of the whole brain.
R. D. Gottapu and C. H. Dagli, "DenseNet for Anatomical Brain Segmentation," Procedia Computer Science, vol. 140, pp. 179-185, Elsevier B.V., Nov 2018.
The definitive version is available at https://doi.org/10.1016/j.procs.2018.10.327
Complex Adaptive Systems Conference with Theme: Cyber Physical Systems and Deep Learning, CAS 2018 (2018: Nov. 5-7, Chicago, IL)
Engineering Management and Systems Engineering
Keywords and Phrases
Convolutional neural network (CNN); Dense Net; Segmentation
International Standard Serial Number (ISSN)
Article - Conference proceedings
© 2019 The Authors, All rights reserved.
Creative Commons Licensing
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.