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

TL;DR

This paper introduces a novel multimodal 3D GAN model with CRF and Pseudo-3D enhancements for accurate brain tumor segmentation, outperforming existing models on the BraTS-2018 dataset.

Contribution

The paper proposes a new 3D-vGAN architecture combining adversarial training, CRF, and Pseudo-3D techniques for improved brain tumor segmentation accuracy.

Findings

Achieves over 99.8% specificity on BraTS-2018 dataset.

Outperforms classical models like U-net, GAN, FCN, and 3D V-net.

Demonstrates the effectiveness of combining adversarial training with CRF and Pseudo-3D methods.

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%.