Department of Nuclear Medicine, The First Affiliated Hospital of Harbin Medical University, Harbin, Heilongjiang, China.
Department of Pathology, The First Affiliated Hospital of Harbin Medical University, Harbin, Heilongjiang, China.
Med Phys. 2024 Aug;51(8):5214-5225. doi: 10.1002/mp.17193. Epub 2024 May 27.
Radiomics has been used in the diagnosis of tumor lymph node metastasis (LNM). However, to date, most studies have been based on intratumoral radiomics. Few studies have focused on the use of F-fluorodeoxyglucose positron emission computed tomography (F-FDG PET/CT) peritumoral radiomics for the diagnosis of LNM in colorectal cancer (CRC).
Determining the value of radiomics features extracted from F-FDG PET/CT images of the peritumoral region in predicting LNM in patients with CRC.
The clinical data and preoperative F-FDG PET/CT images of 244 CRC patients were retrospectively analyzed. Intratumoral and peritumoral radiomics features were screened using the mutual information method, and least absolute shrinkage and selection operator regression. Based on the selected radiomics features, a radiomics score (Rad-score) was calculated, and independent risk factors obtained from univariate and multivariate logistic regression analyses were used to construct clinical and combined (Radiomics + Clinical) models. The performance of these models was evaluated using the DeLong test, while their clinical utility was assessed by decision curve analysis. Finally, a nomogram was constructed to visualize the predictive model.
The most optimal set of features retained by the feature filtering process were all peritumoral radiomic features. Carcinoembryonic antigen levels, PET/CT-reported lymph node status and Rad-score were found to be independent risk factors for LNM. All three LNM risk assessment models exhibited good predictive performance, with the combined model showing the best classification results, with areas under the curve of 0.85 and 0.76 in the training and validation groups, respectively. The DeLong test revealed that the performance of the combined model was superior to that of the clinical and radiomics models in both the training and validation groups, although this difference was only statistically significant in the training group. DCA indicated that the combined model displayed better clinical utility.
F-FDG PET/CT peritumoral radiomics is uniquely suited to predict the presence of LNM in patients with CRC. In particular, the predictive efficacy of LNM for precision therapy and individualized patient management can be improved by using a combination of clinical risk factors.
放射组学已被用于肿瘤淋巴结转移（LNM）的诊断。然而，迄今为止，大多数研究都基于肿瘤内放射组学。很少有研究关注使用 F-氟脱氧葡萄糖正电子发射计算机断层扫描（F-FDG PET/CT）肿瘤周围放射组学来诊断结直肠癌（CRC）中的 LNM。
确定从结直肠癌患者 F-FDG PET/CT 图像肿瘤周围区域提取的放射组学特征在预测 LNM 中的价值。
回顾性分析 244 例 CRC 患者的临床资料和术前 F-FDG PET/CT 图像。使用互信息法和最小绝对收缩和选择算子回归筛选肿瘤内和肿瘤周围放射组学特征。基于所选的放射组学特征，计算放射组学评分（Rad-score），并使用单变量和多变量逻辑回归分析获得独立的风险因素，用于构建临床和联合（放射组学+临床）模型。使用 DeLong 检验评估这些模型的性能，通过决策曲线分析评估其临床实用性。最后，构建一个列线图来可视化预测模型。
特征过滤过程保留的最佳特征集均为肿瘤周围放射组学特征。癌胚抗原水平、PET/CT 报告的淋巴结状态和 Rad-score 被确定为 LNM 的独立危险因素。所有三种 LNM 风险评估模型均表现出良好的预测性能，联合模型的分类结果最好，在训练组和验证组中的曲线下面积分别为 0.85 和 0.76。DeLong 检验表明，在训练组和验证组中，联合模型的性能均优于临床和放射组学模型，尽管这种差异仅在训练组中具有统计学意义。DCA 表明联合模型具有更好的临床实用性。
F-FDG PET/CT 肿瘤周围放射组学特别适合预测 CRC 患者的 LNM 存在。通过结合临床风险因素，可以提高 LNM 对精准治疗和个体化患者管理的预测效果。
