Skett Stephen, Patel Tina, Duprez Didier, Gupta Sunnia, Netherton Tucker, Trauernicht Christoph, Aldridge Sarah, Eaton David, Cardenas Carlos, Court Laurence E, Smith Daniel, Aggarwal Ajay
Guy's and St. Thomas' NHS Foundation Trust, London, United Kingdom.
Stellenbosch University Faculty of Medicine and Health Sciences, Tygerberg Hospital, Cape Town, South Africa.
Phys Imaging Radiat Oncol. 2024 Aug 24;31:100637. doi: 10.1016/j.phro.2024.100637. eCollection 2024 Jul.
In many clinics, positron-emission tomography is unavailable and clinician time extremely limited. Here we describe a deep-learning model for autocontouring gross disease for patients undergoing palliative radiotherapy for primary lung lesions and/or hilar/mediastinal nodal disease, based only on computed tomography (CT) images.
An autocontouring model (nnU-Net) was trained to contour gross disease in 379 cases (352 training, 27 test); 11 further test cases from an external centre were also included. Anchor-point-based post-processing was applied to remove extraneous autocontoured regions. The autocontours were evaluated quantitatively in terms of volume similarity (Dice similarity coefficient [DSC], surface Dice coefficient, 95 percentile Hausdorff distance [HD95], and mean surface distance), and scored for usability by two consultant oncologists. The magnitude of treatment margin needed to account for geometric discrepancies was also assessed.
The anchor point process successfully removed all erroneous regions from the autocontoured disease, and identified two cases to be excluded from further analysis due to 'missed' disease. The average DSC and HD95 were 0.8 ± 0.1 and 10.5 ± 7.3 mm, respectively. A 10-mm uniform margin-distance applied to the autocontoured region was found to yield "full coverage" (sensitivity > 0.99) of the clinical contour for 64 % of cases. Ninety-seven percent of evaluated autocontours were scored by both clinicians as requiring no or minor edits.
Our autocontouring model was shown to produce clinically usable disease outlines, based on CT alone, for approximately two-thirds of patients undergoing lung radiotherapy. Further work is necessary to improve this before clinical implementation.
在许多诊所，正电子发射断层扫描无法使用，且临床医生的时间极为有限。在此，我们描述一种深度学习模型，该模型仅基于计算机断层扫描（CT）图像，用于对原发性肺部病变和/或肺门/纵隔淋巴结疾病接受姑息性放疗的患者的大体疾病进行自动轮廓勾画。
训练一个自动轮廓勾画模型（nnU-Net），以对379例病例（352例训练，27例测试）中的大体疾病进行轮廓勾画；还纳入了来自外部中心的11个额外测试病例。应用基于锚点的后处理来去除多余的自动轮廓勾画区域。根据体积相似性（骰子相似系数[DSC]、表面骰子系数、95百分位数豪斯多夫距离[HD95]和平均表面距离）对自动轮廓进行定量评估，并由两名肿瘤学顾问对可用性进行评分。还评估了考虑几何差异所需的治疗边界大小。
锚点处理成功地从自动轮廓勾画的疾病中去除了所有错误区域，并确定了两例因“遗漏”疾病而需排除在进一步分析之外的病例。平均DSC和HD95分别为0.8±0.1和10.5±7.3毫米。发现对自动轮廓勾画区域应用10毫米的均匀边界距离，在64%的病例中可实现临床轮廓的“完全覆盖”（敏感性>0.99）。97%的评估自动轮廓被两名临床医生评为无需编辑或只需进行少量编辑。
我们的自动轮廓勾画模型被证明仅基于CT就能为大约三分之二接受肺部放疗的患者生成临床可用的疾病轮廓。在临床实施之前，有必要进一步开展工作来改进这一模型。
