深度学习算法在头颈癌PET图像中全自动勾画肿瘤大体体积
本研究评估了3种深度学习算法结合8种损失函数在头颈癌患者PET图像中自动分割肿瘤大体体积(GTV)的性能,使用470例患者的18F-FDG PET/CT图像进行训练、评估和测试。结果显示,Res-Net结合交叉熵损失、Dense-VNet和NN-UNet结合Dice加XEnt损失获得了较高的Dice系数(0.86-0.87),且SUVmax量化的相对误差较低(<5%)。作者认为深度学习算法在自动GTV勾画方面表现出良好前景,但临床部署需谨慎对待异常值。
要点
- Res-Net结合交叉熵损失获得最佳Dice系数0.86±0.05(95%CI 0.85-0.87)。
- 三种网络间Dice系数差异无统计学意义(P>0.05)。
- Dice系数>0.84时SUVmax量化相对误差<5%,Res-Net交叉熵损失达0.41%。
核心结论
- 深度学习算法在头颈癌PET图像自动GTV勾画中表现出有前景的性能。
- 不同损失函数与不同网络组合时性能具有竞争力,Res-Net结合交叉熵损失、Dense-VNet和NN-UNet结合Dice加XEnt损失成为可靠的GTV勾画网络。
- 由于深度学习算法中出现异常值,临床部署需谨慎。
研究设计与样本
- 使用470例头颈癌患者的18F-FDG PET/CT图像,手动勾画的GTV作为参考标准。
- 训练集340例,评估集30例,测试集100例来自不同中心。
- 构建了28种不同的网络组合:3种网络(Dense-VNet、NN-UNet、Res-Net)与8种损失函数(Dice、广义Wasserstein Dice损失、Dice加XEnt损失、广义Dice损失、交叉熵、敏感度-特异度、Tversky等)组合。
- 采用旋转(-15度,+15度)、缩放(-20%,20%)、随机翻转(3轴)和弹性变形(sigma=1,变形比例=0.7)进行数据增强。
- 使用标准图像分割指标(Dice相似度)、图像衍生PET指标、一阶和形状放射组学特征评估性能。
主要发现
- 最佳Dice系数(均值±标准差)由交叉熵用于Res-Net(0.86±0.05;95%CI 0.85-0.87)、Dense-VNet(0.85±0.058;95%CI 0.84-0.86)和Dice加XEnt用于NN-UNet(0.87±0.05;95%CI 0.86-0.88)获得。
- 3种网络之间的差异无统计学意义(P>0.05)。
- 在Dice系数大于0.84的网络中,SUVmax量化的相对误差(RE%)小于5%,而Res-Net结合交叉熵损失实现了更低的RE%(0.41%)。
- 对于最大三维直径和球形度形状特征,所有网络的相对误差分别≤5%和≤10%,反映变异性较小。
关键数值与推荐
- 最佳Dice系数:Res-Net结合交叉熵损失0.86±0.05(95%CI 0.85-0.87),Dense-VNet 0.85±0.058(95%CI 0.84-0.86),NN-UNet结合Dice加XEnt损失0.87±0.05(95%CI 0.86-0.88)。
- 3种网络间Dice系数差异无统计学意义(P>0.05)。
- Dice系数>0.84时SUVmax量化相对误差<5%,Res-Net结合交叉熵损失RE%为0.41%。
- 最大三维直径相对误差≤5%,球形度相对误差≤10%。
临床怎么用
- 深度学习算法可用于头颈癌PET图像中GTV的自动勾画,有望辅助临床工作流程。
- Res-Net结合交叉熵损失、Dense-VNet和NN-UNet结合Dice加XEnt损失可作为可靠的GTV勾画网络选择。
- 临床部署时需谨慎,注意深度学习算法可能出现的异常值。
局限性
- 深度学习算法在临床部署中可能出现异常值,需谨慎对待。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
PURPOSE: The availability of automated, accurate, and robust gross tumor volume (GTV) segmentation algorithms is critical for the management of head and neck cancer (HNC) patients. In this work, we evaluated 3 state-of-the-art deep learning algorithms combined with 8 different loss functions for PET image segmentation using a comprehensive training set and evaluated its performance on an external validation set of HNC patients. PATIENTS AND METHODS: 18F-FDG PET/CT images of 470 patients presenting with HNC on which manually defined GTVs serving as standard of reference were used for training (340 patients), evaluation (30 patients), and testing (100 patients from different centers) of these algorithms. PET image intensity was converted to SUVs and normalized in the range (0-1) using the SUVmax of the whole data set. PET images were cropped to 12 × 12 × 12 cm3 subvolumes using isotropic voxel spacing of 3 × 3 × 3 mm3 containing the whole tumor and neighboring background including lymph nodes. We used different approaches for data augmentation, including rotation (-15 degrees, +15 degrees), scaling (-20%, 20%), random flipping (3 axes), and elastic deformation (sigma = 1 and proportion to deform = 0.7) to increase the number of training sets. Three state-of-the-art networks, including Dense-VNet, NN-UNet, and Res-Net, with 8 different loss functions, including Dice, generalized Wasserstein Dice loss, Dice plus XEnt loss, generalized Dice loss, cross-entropy, sensitivity-specificity, and Tversky, were used. Overall, 28 different networks were built. Standard image segmentation metrics, including Dice similarity, image-derived PET metrics, first-order, and shape radiomic features, were used for performance assessment of these algorithms. RESULTS: The best results in terms of Dice coefficient (mean ± SD) were achieved by cross-entropy for Res-Net (0.86 ± 0.05; 95% confidence interval [CI], 0.85-0.87), Dense-VNet (0.85 ± 0.058; 95% CI, 0.84-0.86), and Dice plus XEnt for NN-UNet (0.87 ± 0.05; 95% CI, 0.86-0.88). The difference between the 3 networks was not statistically significant (P > 0.05). The percent relative error (RE%) of SUVmax quantification was less than 5% in networks with a Dice coefficient more than 0.84, whereas a lower RE% (0.41%) was achieved by Res-Net with cross-entropy loss. For maximum 3-dimensional diameter and sphericity shape features, all networks achieved a RE ≤ 5% and ≤10%, respectively, reflecting a small variability. CONCLUSIONS: Deep learning algorithms exhibited promising performance for automated GTV delineation on HNC PET images. Different loss functions performed competitively when using different networks and cross-entropy for Res-Net, Dense-VNet, and Dice plus XEnt for NN-UNet emerged as reliable networks for GTV delineation. Caution should be exercised for clinical deployment owing to the occurrence of outliers in deep learning-based algorithms.
引用
- Shiri I, Arabi H, Sanaat A, et al. Fully Automated Gross Tumor Volume Delineation From PET in Head and Neck Cancer Using Deep Learning Algorithms. Clin Nucl Med. 2021. PMID: 34238799. doi:10.1097/RLU.0000000000003789
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。