Multimodal 3D Brain Tumor Segmentation with Adversarial Training and Conditional Random Field

TL;DR

Results show that 3D-vGAN outperforms classical segmentation models, including U-net, Gan, FCN and 3D V-net, reaching specificity over 99.8%.

Abstract

Accurate brain tumor segmentation remains a challenging task due to structural complexity and great individual differences of gliomas. Leveraging the pre-eminent detail resilience of CRF and spatial feature extraction capacity of V-net, we propose a multimodal 3D Volume Generative Adversarial Network (3D-vGAN) for precise segmentation. The model utilizes Pseudo-3D for V-net improvement, adds conditional random field after generator and use original image as supplemental guidance. Results, using the BraTS-2018 dataset, show that 3D-vGAN outperforms classical segmentation models, including U-net, Gan, FCN and 3D V-net, reaching specificity over 99.8%.

Figures (7)

• Figure 1: 3D-vGAN overall network architecture.
• Figure 2: The model structure of the Generator.
• Figure 3: The model structure of the Discriminator cirillo2021.
• Figure 4: Flow Chart of Conditional Random Fields.
• Figure 5: a) Training and validation loss b) dice score curves.