Radiation Oncology Department, University Hospital, Brest, France.
LaTIM, INSERM, UMR 1101, Univ Brest, Brest, France.
Eur J Nucl Med Mol Imaging. 2023 Jul;50(8):2514-2528. doi: 10.1007/s00259-023-06180-w. Epub 2023 Mar 9.
To develop machine learning models to predict para-aortic lymph node (PALN) involvement in patients with locally advanced cervical cancer (LACC) before chemoradiotherapy (CRT) using F-FDG PET/CT and MRI radiomics combined with clinical parameters.
We retrospectively collected 178 patients (60% for training and 40% for testing) in 2 centers and 61 patients corresponding to 2 further external testing cohorts with LACC between 2010 to 2022 and who had undergone pretreatment analog or digital F-FDG PET/CT, pelvic MRI and surgical PALN staging. Only primary tumor volumes were delineated. Radiomics features were extracted using the Radiomics toolbox®. The ComBat harmonization method was applied to reduce the batch effect between centers. Different prediction models were trained using a neural network approach with either clinical, radiomics or combined models. They were then evaluated on the testing and external validation sets and compared.
In the training set (n = 102), the clinical model achieved a good prediction of the risk of PALN involvement with a C-statistic of 0.80 (95% CI 0.71, 0.87). However, it performed in the testing (n = 76) and external testing sets (n = 30 and n = 31) with C-statistics of only 0.57 to 0.67 (95% CI 0.36, 0.83). The ComBat-radiomic (GLDZM_HISDE_PET_FBN64 and Shape_maxDiameter2D3_PET_FBW0.25) and ComBat-combined (FIGO 2018 and same radiomics features) models achieved very high predictive ability in the training set and both models kept the same performance in the testing sets, with C-statistics from 0.88 to 0.96 (95% CI 0.76, 1.00) and 0.85 to 0.92 (95% CI 0.75, 0.99), respectively.
Radiomic features extracted from pre-CRT analog and digital F-FDG PET/CT outperform clinical parameters in the decision to perform a para-aortic node staging or an extended field irradiation to PALN. Prospective validation of our models should now be carried out.
利用 F-FDG PET/CT 和 MRI 放射组学结合临床参数，为接受放化疗（CRT）前局部晚期宫颈癌（LACC）患者开发预测腹膜后淋巴结（PALN）受累的机器学习模型。
我们回顾性收集了 2010 年至 2022 年期间在 2 个中心接受治疗的 178 名患者（60%用于训练，40%用于测试）和另外 61 名患者（分别对应 2 个外部测试队列），这些患者均患有 LACC，并进行了预处理模拟或数字 F-FDG PET/CT、盆腔 MRI 和手术 PALN 分期。仅对原发肿瘤体积进行了勾画。使用 Radiomics 工具框®提取放射组学特征。应用 ComBat 均衡化方法减少中心之间的批次效应。使用神经网络方法分别使用临床、放射组学或联合模型训练不同的预测模型，然后在测试集和外部验证集上进行评估并进行比较。
在训练集（n=102）中，临床模型对 PALN 受累风险的预测具有良好的效果，C 统计量为 0.80（95%CI 0.71，0.87）。然而，在测试集（n=76）和外部测试集（n=30 和 n=31）中，其 C 统计量仅为 0.57 至 0.67（95%CI 0.36，0.83）。ComBat-radiomic（GLDZM_HISDE_PET_FBN64 和 Shape_maxDiameter2D3_PET_FBW0.25）和 ComBat-联合（FIGO 2018 和相同的放射组学特征）模型在训练集中具有非常高的预测能力，并且在测试集中保持相同的性能，C 统计量分别为 0.88 至 0.96（95%CI 0.76，1.00）和 0.85 至 0.92（95%CI 0.75，0.99）。
在决定是否进行主动脉旁淋巴结分期或对 PALN 进行扩展野照射时，预处理模拟和数字 F-FDG PET/CT 提取的放射组学特征优于临床参数。现在应该进行我们模型的前瞻性验证。
