Department of Gynecological Oncology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China.
Department of Gynecological Oncology, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, Guangzhou, China
Int J Gynecol Cancer. 2024 Sep 2;34(9):1437-1444. doi: 10.1136/ijgc-2024-005580.
To predict preoperative inguinal lymph node metastasis in vulvar cancer patients using a machine learning model based on imaging features and clinical data from pelvic magnetic resonance imaging (MRI).
52 vulvar cancer patients were divided into a training set (n=37) and validation set (n=15). Clinical data and MRI images were collected, and regions of interest were delineated by experienced radiologists. A total of 1688 quantitative imaging features were extracted using the Radcloud platform. Dimensionality reduction and feature selection were applied, resulting in a radiomics signature. Clinical characteristics were screened, and a combined model integrating the radiomics signature and significant clinical features was constructed using logistic regression. Four machine learning classifiers (K nearest neighbor, random forest, adaptive boosting, and latent dirichlet allocation) were trained and validated. Model performance was evaluated using the receiver operating characteristic curve and the area under the curve (AUC), as well as decision curve analysis.
The radiomics score significantly differentiated between lymph node metastasis positive and negative patients in both the training and validation sets. The combined model demonstrated excellent discrimination, with AUC values of 0.941 and 0.933 in the training and validation sets, respectively. The calibration curve and decision curve analysis confirmed the model's high predictive accuracy and clinical utility. Among the machine learning classifiers, latent dirichlet allocation and random forest models achieved AUC values >0.7 in the validation set. Integrating all four classifiers resulted in a total model with an AUC of 0.717 in the validation set.
Radiomics combined with artificial intelligence can provide a new method for prediction of inguinal lymph node metastasis of vulvar cancer before surgery.
利用基于盆腔磁共振成像（MRI）影像特征和临床数据的机器学习模型预测外阴癌患者术前腹股沟淋巴结转移。
将 52 例外阴癌患者分为训练集（n=37）和验证集（n=15）。收集临床数据和 MRI 图像，由有经验的放射科医生勾画感兴趣区。使用 Radcloud 平台提取 1688 个定量影像特征。进行降维和特征选择，得到放射组学特征。筛选临床特征，利用逻辑回归构建整合放射组学特征和显著临床特征的联合模型。使用 K 最近邻、随机森林、自适应增强和潜在狄利克雷分配这四种机器学习分类器进行训练和验证。使用受试者工作特征曲线和曲线下面积（AUC）以及决策曲线分析评估模型性能。
放射组学评分在训练集和验证集均能显著区分淋巴结转移阳性和阴性患者。联合模型具有出色的判别能力，在训练集和验证集的 AUC 值分别为 0.941 和 0.933。校准曲线和决策曲线分析证实了模型具有较高的预测准确性和临床实用性。在机器学习分类器中，潜在狄利克雷分配和随机森林模型在验证集的 AUC 值均大于 0.7。整合所有四个分类器得到的总模型在验证集的 AUC 值为 0.717。
放射组学结合人工智能可以为外阴癌术前预测腹股沟淋巴结转移提供一种新方法。
