[论文解读] Hierarchical 3D fully convolutional networks for multi-organ segmentation

Holger R. Roth, Hirohisa Oda+6 more|arXiv (Cornell University)|Apr 21, 2017

Advanced Neural Network Applications参考文献 12被引用 119

一句话总结

该论文提出一个两阶段粗到细的3D FCN（基于3D U-Net），用于CT扫描的多器官分割，在每个器官的平均Dice提高7.5个百分点，并在未见数据上表现强劲。

ABSTRACT

Recent advances in 3D fully convolutional networks (FCN) have made it feasible to produce dense voxel-wise predictions of full volumetric images. In this work, we show that a multi-class 3D FCN trained on manually labeled CT scans of seven abdominal structures (artery, vein, liver, spleen, stomach, gallbladder, and pancreas) can achieve competitive segmentation results, while avoiding the need for handcrafting features or training organ-specific models. To this end, we propose a two-stage, coarse-to-fine approach that trains an FCN model to roughly delineate the organs of interest in the first stage (seeing $\sim$40% of the voxels within a simple, automatically generated binary mask of the patient's body). We then use these predictions of the first-stage FCN to define a candidate region that will be used to train a second FCN. This step reduces the number of voxels the FCN has to classify to $\sim$10% while maintaining a recall high of $>$99%. This second-stage FCN can now focus on more detailed segmentation of the organs. We respectively utilize training and validation sets consisting of 281 and 50 clinical CT images. Our hierarchical approach provides an improved Dice score of 7.5 percentage points per organ on average in our validation set. We furthermore test our models on a completely unseen data collection acquired at a different hospital that includes 150 CT scans with three anatomical labels (liver, spleen, and pancreas). In such challenging organs as the pancreas, our hierarchical approach improves the mean Dice score from 68.5 to 82.2%, achieving the highest reported average score on this dataset.

研究动机与目标

• 在不使用器官特定模型的情况下，推动CT中准确的多器官分割。
• 展示分层的粗到细FCN能够聚焦于困难边界。
• 展示跨不同医院数据集的泛化能力。
• 量化七个腹部结构的Dice分数改进。

提出的方法

• 使用3D U-Net FCN对八个类别进行分割（七个器官加背景）。
• 分两阶段训练：阶段1大致界定一个身体尺度的候选区域（约占体素的40%）。
• 阶段2在更小的候选区域内细化分割（约占体素的10%）。
• 对输入进行下采样以增大视野，并在测试阶段应用滑动拼接策略。
• 应用逐体素交叉熵损失并带有类别权重，以平衡前景和背景体素。
• 使用随机3D弹性形变和随机旋转来增强数据。

实验结果

研究问题

• RQ1 单个3D FCN模型是否可以在无需器官特异模型的情况下分割多种腹部器官？
• RQ2 分层的粗到细方法是否提高分割准确性，尤其是对小而细的结构？
• RQ3 该方法对来自不同医院和扫描仪的未见数据的泛化能力如何？

主要发现

• 在验证集上，每个器官的平均Dice分数提升7.5个百分点（Stage 2 相对于 Stage 1）。
• 阶段1的召回率>99%，在将候选区域扩张到r=3时大约有10%的假阳性。
• 动脉在阶段2中Dice从59.0提升到79.6；胰腺在阶段2中Dice从54.8提升到63.1。
• 在未见测试数据（肝、脾、胰腺）上，非重叠阶段2在胰腺的平均Dice为68.5，肝为93.2，脾为89.7；重叠拼接将胰腺提高到82.2 Dice。
• 该方法在未见数据集上取得了接近最先进的结果，肝和胰腺的性能受益于分层方法。