[论文解读] VoxResNet: Deep Voxelwise Residual Networks for Volumetric Brain Segmentation

Hao Chen, Qi Dou+2 more|arXiv (Cornell University)|Aug 21, 2016

Advanced Neural Network Applications参考文献 34被引用 115

一句话总结

VoxResNet 将残差学习扩展到 3D，用于体积脑分割，并通过融合多模态数据和自上下文信息来提升性能。

ABSTRACT

Recently deep residual learning with residual units for training very deep neural networks advanced the state-of-the-art performance on 2D image recognition tasks, e.g., object detection and segmentation. However, how to fully leverage contextual representations for recognition tasks from volumetric data has not been well studied, especially in the field of medical image computing, where a majority of image modalities are in volumetric format. In this paper we explore the deep residual learning on the task of volumetric brain segmentation. There are at least two main contributions in our work. First, we propose a deep voxelwise residual network, referred as VoxResNet, which borrows the spirit of deep residual learning in 2D image recognition tasks, and is extended into a 3D variant for handling volumetric data. Second, an auto-context version of VoxResNet is proposed by seamlessly integrating the low-level image appearance features, implicit shape information and high-level context together for further improving the volumetric segmentation performance. Extensive experiments on the challenging benchmark of brain segmentation from magnetic resonance (MR) images corroborated the efficacy of our proposed method in dealing with volumetric data. We believe this work unravels the potential of 3D deep learning to advance the recognition performance on volumetric image segmentation.

研究动机与目标

• 通过在 3D 中使用深度残差学习，推动鲁棒的体积脑分割。
• 提出 VoxResNet，一个用于体积分割的深度体素级残差网络。
• 引入自上下文 VoxResNet，整合低层、中层和高层上下文信息。
• Demonstrate benefits of multi-modality data fusion in MR brain segmentation.

提出的方法

• 将 2D 深度残差网络扩展为具有 25 个体积卷积层和 4 个上采样卷积层的 3D VoxResNet。
• 使用步长为 2 的 3x3x3 内核以实现较大感受野。
• 纳入带跳跃连接的后激活残差单元，以实现稳定的深度训练。
• 通过拼接和联合学习整合多模态输入（如 T1、T1-IR、T2-FLAIR）。
• 应用四个辅助分类器（C1–C4）进行深度监督以融合多尺度上下文。
• 采用自上下文方案，其中初始 VoxResNet 输出引导第二阶段的 Auto-context VoxResNet 进行细化。
• 训练时使用结合正则化和体素级交叉熵的损失，并包括带加权强调的辅助损失。

实验结果

研究问题

• RQ1 3D 深度残差网络（VoxResNet）是否能够有效从 MR 图像中学习体积脑分割？
• RQ2 多模态数据融合是否在 3D 情境下提高分割准确性？
• RQ3 整合自上下文信息是否进一步提升分割性能？
• RQ4 深度监督对 3D 体积分割准确性有何实际影响？

主要发现

• VoxResNet 结合多模态输入在脑组织分割方面优于单模态基线。
• 自上下文集成进一步提升 Dice 系数 (DC) 并减少分割误差。
• All 模态设置在脑组织（GM、WM、CSF）上实现较高 DC 值，且 Hausdorff 距离和绝对体积差异具有竞争力。
• Auto-context VoxResNet 相对于单独的 VoxResNet，在验证测试中获得额外增益。
• 在 MRBrainS 基准测试中，所提出的方法位列前列，显示出体积脑分割的强劲性能。