Department of Radiation Oncology, Washington University in St Louis School of Medicine, St. Louis, Missouri, USA.
Med Phys. 2021 Nov;48(11):7172-7188. doi: 10.1002/mp.15237. Epub 2021 Sep 30.
To develop and evaluate deep learning-based autosegmentation of cardiac substructures from noncontrast planning computed tomography (CT) images in patients undergoing breast cancer radiotherapy and to investigate the algorithm sensitivity to out-of-distribution data such as CT image artifacts.
Nine substructures including aortic valve (AV), left anterior descending (LAD), tricuspid valve (TV), mitral valve (MV), pulmonic valve (PV), right atrium (RA), right ventricle (RV), left atrium (LA), and left ventricle (LV) were manually delineated by a radiation oncologist on noncontrast CT images of 129 patients with breast cancer; among them 90 were considered in-distribution data, also named as "clean" data. The image/label pairs of 60 subjects were used to train a 3D deep neural network while the remaining 30 were used for testing. The rest of the 39 patients were considered out-of-distribution ("outlier") data, which were used to test robustness. Random rigid transformations were used to augment the dataset during training. We investigated multiple loss functions, including Dice similarity coefficient (DSC), cross-entropy (CE), Euclidean loss as well as the variation and combinations of these, data augmentation, and network size on overall performance and sensitivity to image artifacts due to infrequent events such as the presence of implanted devices. The predicted label maps were compared to the ground-truth labels via DSC and mean and 90th percentile symmetric surface distance (90th-SSD).
When modified Dice combined with cross-entropy (MD-CE) was used as the loss function, the algorithm achieved a mean DSC = 0.79 ± 0.07 for chambers and 0.39 ± 0.10 for smaller substructures (valves and LAD). The mean and 90th-SSD were 2.7 ± 1.4 and 6.5 ± 2.8 mm for chambers and 4.1 ± 1.7 and 8.6 ± 3.2 mm for smaller substructures. Models with MD-CE, Dice-CE, MD, and weighted CE loss had highest performance, and were statistically similar. Data augmentation did not affect model performances on both clean and outlier data and model robustness was susceptible to network size. For a certain type of outlier data, robustness can be improved via incorporating them into the training process. The execution time for segmenting each patient was on an average 2.1 s.
A deep neural network provides a fast and accurate segmentation of large cardiac substructures in noncontrast CT images. Model robustness of two types of clinically common outlier data were investigated and potential approaches to improve them were explored. Evaluation of clinical acceptability and integration into clinical workflow are pending.
开发并评估基于深度学习的乳腺癌放疗患者非对比规划 CT 图像心脏亚结构自动分割算法，并研究算法对 CT 图像伪影等分布外数据的敏感性。
在 129 例乳腺癌患者的非对比 CT 图像上，由一名放射肿瘤学家手动勾画 9 个亚结构，包括主动脉瓣（AV）、左前降支（LAD）、三尖瓣（TV）、二尖瓣（MV）、肺动脉瓣（PV）、右心房（RA）、右心室（RV）、左心房（LA）和左心室（LV）；其中 90 例为分布内数据，也称为“干净”数据。利用 60 例患者的图像/标签对训练 3D 深度神经网络，其余 30 例用于测试。其余 39 例患者被认为是分布外（“异常”）数据，用于测试鲁棒性。在训练过程中，使用随机刚体变换来扩充数据集。我们研究了多种损失函数，包括 Dice 相似系数（DSC）、交叉熵（CE）、欧几里得损失以及它们的变体和组合、数据扩充和网络规模对整体性能的影响，以及对由于植入设备等罕见事件引起的图像伪影的敏感性。通过 DSC 和平均和 90 百分位对称面距离（90th-SSD），将预测的标签图与地面真实标签进行比较。
当使用改进的 Dice 结合交叉熵（MD-CE）作为损失函数时，该算法在腔室方面的平均 DSC 为 0.79±0.07，在较小亚结构（瓣膜和 LAD）方面的平均 DSC 为 0.39±0.10。腔室的平均和 90th-SSD 分别为 2.7±1.4mm 和 6.5±2.8mm，较小亚结构的平均和 90th-SSD 分别为 4.1±1.7mm 和 8.6±3.2mm。具有 MD-CE、Dice-CE、MD 和加权 CE 损失的模型具有最高的性能，且统计学上无显著差异。数据扩充对清洁数据和异常数据的模型性能没有影响，模型鲁棒性对网络规模敏感。对于某种类型的异常数据，可以通过将其纳入训练过程来提高模型的鲁棒性。每个患者的分割执行时间平均为 2.1s。
深度神经网络可快速准确地分割非对比 CT 图像中的大型心脏亚结构。研究了两种常见临床异常数据的模型鲁棒性，并探讨了提高其鲁棒性的潜在方法。临床可接受性评估和整合到临床工作流程有待进一步研究。
