Coronary calcification is a strong indicator of coronary artery disease and a key determinant of the outcome of percutaneous coronary intervention. We propose a fully automated method to segment and quantify coronary calcification in intravascular OCT (IVOCT) images based on convolutional neural networks (CNN). All possible calcified plaques were segmented from IVOCT pullbacks using a spatial-temporal encoder-decoder network by exploiting the 3D continuity information of the plaques, which were then screened and classified by a DenseNet network to reduce false positives. A novel data augmentation method based on the IVOCT image acquisition pattern was also proposed to improve the performance and robustness of the segmentation. Clinically relevant metrics including calcification area, depth, angle, thickness, volume, and stent-deployment calcification score, were automatically computed. 13844 IVOCT images with 2627 calcification slices from 45 clinical OCT pullbacks were collected and used to train and test the model. The proposed method performed significantly better than existing state-of-the-art 2D and 3D CNN methods. The data augmentation method improved the Dice similarity coefficient for calcification segmentation from 0.615±0.332 to 0.756±0.222, reaching human-level inter-observer agreement. Our proposed region-based classifier improved image-level calcification classification precision and F1-score from 0.725±0.071 and 0.791±0.041 to 0.964±0.002 and 0.883±0.008, respectively. Bland–Altman analysis showed close agreement between manual and automatic calcification measurements. Our proposed method is valuable for automated assessment of coronary calcification lesions and in-procedure planning of stent deployment.

中文翻译：

---

使用时空编码器-解码器网络综合评估血管内 OCT 中的冠状动脉钙化

冠状动脉钙化是冠状动脉疾病的重要指标，也是经皮冠状动脉介入治疗结果的关键决定因素。我们提出了一种基于卷积神经网络 (CNN) 的全自动方法来分割和量化血管内 OCT (IVOCT) 图像中的冠状动脉钙化。通过利用斑块的 3D 连续性信息，使用时空编码器-解码器网络从 IVOCT 回调中分割所有可能的钙化斑块，然后通过 DenseNet 网络对其进行筛选和分类以减少误报。还提出了一种基于 IVOCT 图像采集模式的新型数据增强方法，以提高分割的性能和鲁棒性。临床相关指标，包括钙化面积、深度、角度、厚度、体积、自动计算支架部署钙化评分。收集了来自 45 个临床 OCT 回调的 13844 张 IVOCT 图像和 2627 个钙化切片，用于训练和测试模型。所提出的方法的性能明显优于现有的最先进的 2D 和 3D CNN 方法。数据增强方法将钙化分割的Dice相似系数从0.615±0.332提高到0.756±0.222，达到人类水平的观察者间一致性。我们提出的基于区域的分类器将图像级钙化分类精度和 F1 分数分别从 0.725±0.071 和 0.791±0.041 提高到 0.964±0.002 和 0.883±0.008。Bland-Altman 分析显示手动和自动钙化测量值非常一致。