Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.
The University of Texas MD Anderson Graduate School of Biomedical Science, Houston, TX, USA.
Med Phys. 2021 Apr;48(4):1697-1706. doi: 10.1002/mp.14728. Epub 2021 Feb 19.
To enable generation of high-quality deep learning segmentation models from severely limited contoured cases (e.g., ~10 cases).
Thirty head and neck computed tomography (CT) scans with well-defined contours were deformably registered to 200 CT scans of the same anatomic site without contours. Acquired deformation vector fields were used to train a principal component analysis (PCA) model for each of the 30 contoured CT scans by capturing the mean deformation and most prominent variations. Each PCA model can produce an infinite number of synthetic CT scans and corresponding contours by applying random deformations. We used 300, 600, 1000, and 2000 synthetic CT scans and contours generated from one PCA model to train V-Net, a 3D convolutional neural network architecture, to segment parotid and submandibular glands. We repeated the training using same numbers of training cases generated from 7, 10, 20, and 30 PCA models, with the data distributed evenly between each PCA model. Performance of the segmentation models was evaluated with Dice similarity coefficients between auto-generated contours and physician-drawn contours on 162 test CT scans for parotid glands and another 21 test CT scans for submandibular glands.
Dice values varied with the number of synthetic CT scans and the number of PCA models used to train the network. By using 2000 synthetic CT scans generated from 10 PCA models, we achieved Dice values of 82.8% ± 6.8% for right parotid, 82.0% ± 6.9% for left parotid, and 74.2% ± 6.8% for submandibular glands. These results are comparable with those obtained from state-of-the-art auto-contouring approaches, including a deep learning network trained from more than 1000 contoured patients and a multi-atlas algorithm from 12 well-contoured atlases. Improvement was marginal when >10 PCA models or >2000 synthetic CT scans were used.
We demonstrated an effective data augmentation approach to train high-quality deep learning segmentation models from a limited number of well-contoured patient cases.
从严重受限的勾画病例（例如，约 10 个病例）中生成高质量的深度学习分割模型。
30 例头颈部 CT 扫描具有明确的轮廓，可变形地注册到 200 例具有相同解剖部位但无轮廓的 CT 扫描。获得的变形矢量场用于为 30 例勾画 CT 扫描中的每一例训练主成分分析（PCA）模型，以捕获平均变形和最突出的变化。每个 PCA 模型都可以通过应用随机变形从一个 PCA 模型生成无限数量的合成 CT 扫描和相应的轮廓。我们使用了 300、600、1000 和 2000 个从一个 PCA 模型生成的合成 CT 扫描和轮廓来训练 V-Net，这是一种 3D 卷积神经网络架构，用于分割腮腺和下颌下腺。我们使用相同数量的从 7、10、20 和 30 个 PCA 模型生成的训练病例重复了训练，数据在每个 PCA 模型之间均匀分布。在 162 例腮腺和 21 例下颌下腺的测试 CT 扫描上，通过自动生成轮廓与医生绘制轮廓之间的 Dice 相似系数来评估分割模型的性能。
Dice 值随合成 CT 扫描的数量和用于训练网络的 PCA 模型数量而变化。通过使用从 10 个 PCA 模型生成的 2000 个合成 CT 扫描，我们获得了右侧腮腺 82.8%±6.8%、左侧腮腺 82.0%±6.9%和下颌下腺 74.2%±6.8%的 Dice 值。这些结果与最先进的自动勾画方法相当，包括从 1000 多个勾画患者中训练的深度学习网络和从 12 个良好勾画图谱中使用的多图谱算法。当使用>10 个 PCA 模型或>2000 个合成 CT 扫描时，改善效果很微小。
我们展示了一种有效的数据增强方法，可以从有限数量的勾画良好的患者病例中训练高质量的深度学习分割模型。
