Fully automatic segmentation of diffuse large B cell lymphoma lesions on 3D FDG-PET/CT for total metabolic tumour volume prediction using a convolutional neural network.
Fully automatic segmentation of diffuse large B cell lymphoma lesions on 3D FDG-PET/CT for total metabolic tumour volume prediction using a convolutional neural network.
复制标题
DOI：
10.1007/s00259-020-05080-7
复制
发表时间：
2020-10-24
影响因子：
9.1
通讯作者：
Itti, Emmanuel
中科院分区：
文献类型：
--
作者：
Blanc-Durand, Paul;Jegou, Simon;Itti, Emmanuel
目的 由于涉及的节点、器官或生理摄取的多样性，全身 FDG-PET/CT 上的淋巴瘤病灶检测和分割是一项具有挑战性的任务。我们试图研究三维 (3D) 卷积神经网络 (CNN) 在弥漫性大 B 细胞淋巴瘤 (DLBCL) 患者的大型数据集中自动分割总代谢肿瘤体积 (TMTV) 的性能。方法 数据集包含来自 2 项前瞻性淋巴瘤研究协会 (LYSA) 试验的 733 名 DLBCL 患者的治疗前 FDG-PET/CT。第一个队列 (n = 639) 使用 5 倍交叉验证方案进行训练。第二组 (n = 94) 用于 TMTV 预测的外部验证。对淋巴瘤病变进行 41% SUVmax 自适应阈值处理后，手动获得地面实况掩模。具有 2 个用于 PET 和 CT 的输入通道的 3D U 网架构在 PET/CT 内随机采样的补丁上进行训练，并具有求和交叉熵和 Dice 相似系数 (DSC) 损失。分割性能通过 DSC 和 Jaccard 系数进行评估。最后，TMTV 预测在第二个独立队列中得到验证。结果 验证集中的平均 DSC 和 Jaccard 系数（+/- 标准差）分别为 0.73 +/- 0.20 和 0.68 +/- 0.21。在第一组的验证集中发现平均 TMTV 被低估了 - 12 mL (2.8%) +/- 263 (P = 0.27)。在第二组中，平均 TMTV 低估了 - 116 mL (20.8%) +/- 425，具有统计学意义 (P = 0.01)。结论 我们的 CNN 是一种很有前途的自动检测和分割淋巴瘤病变的工具，尽管 TMTV 被低估了。该 CNN 的全自动和开源功能将有助于提高常规实践中的传播以及淋巴瘤患者 TMTV 评估的可重复性。
Purpose Lymphoma lesion detection and segmentation on whole-body FDG-PET/CT are a challenging task because of the diversity of involved nodes, organs or physiological uptakes. We sought to investigate the performances of a three-dimensional (3D) convolutional neural network (CNN) to automatically segment total metabolic tumour volume (TMTV) in large datasets of patients with diffuse large B cell lymphoma (DLBCL). Methods The dataset contained pre-therapy FDG-PET/CT from 733 DLBCL patients of 2 prospective LYmphoma Study Association (LYSA) trials. The first cohort (n = 639) was used for training using a 5-fold cross validation scheme. The second cohort (n = 94) was used for external validation of TMTV predictions. Ground truth masks were manually obtained after a 41% SUVmax adaptive thresholding of lymphoma lesions. A 3D U-net architecture with 2 input channels for PET and CT was trained on patches randomly sampled within PET/CTs with a summed cross entropy and Dice similarity coefficient (DSC) loss. Segmentation performance was assessed by the DSC and Jaccard coefficients. Finally, TMTV predictions were validated on the second independent cohort. Results Mean DSC and Jaccard coefficients (+/- standard deviation) in the validations set were 0.73 +/- 0.20 and 0.68 +/- 0.21, respectively. An underestimation of mean TMTV by - 12 mL (2.8%) +/- 263 was found in the validation sets of the first cohort (P = 0.27). In the second cohort, an underestimation of mean TMTV by - 116 mL (20.8%) +/- 425 was statistically significant (P = 0.01). Conclusion Our CNN is a promising tool for automatic detection and segmentation of lymphoma lesions, despite slight underestimation of TMTV. The fully automatic and open-source features of this CNN will allow to increase both dissemination in routine practice and reproducibility of TMTV assessment in lymphoma patients.