Radiomics analysis for prediction of lymph node metastasis after neoadjuvant chemotherapy based on pretreatment MRI in patients with locally advanced cervical cancer
Jinjin Liu,Linxiao Dong,Xiaoxian Zhang,Qingxia Wu,Zihan Yang,Yuejie Zhang,Chunmiao Xu,Qingxia Wu,Meiyun Wang
DOI: https://doi.org/10.3389/fonc.2024.1376640
IF: 4.7
2024-05-08
Frontiers in Oncology
Abstract:Background: This study aims to develop and validate a pretreatment MRI-based radiomics model to predict lymph node metastasis (LNM) following neoadjuvant chemotherapy (NACT) in patients with locally advanced cervical cancer (LACC). Methods: Patients with LACC who underwent NACT from two centers between 2013 and 2022 were enrolled retrospectively. Based on the lymph node (LN) status determined in the pathology reports after radical hysterectomy, patients were categorized as LN positive or negative. The patients from center 1 were assigned as the training set while those from center 2 formed the validation set. Radiomics features were extracted from pretreatment sagittal T2-weighted imaging (Sag-T2WI), axial diffusion-weighted imaging (Ax-DWI), and the delayed phase of dynamic contrast-enhanced sagittal T1-weighted imaging (Sag-T1C) for each patient. The K-best and least absolute shrinkage and selection operator (LASSO) methods were employed to reduce dimensionality, and the radiomics features strongly associated with LNM were selected and used to construct three single-sequence models. Furthermore, clinical variables were incorporated through multivariate regression analysis and fused with the selected radiomics features to construct the clinical-radiomics combined model. The diagnostic performance of the models was assessed using receiver operating characteristic (ROC) curve analysis. The clinical utility of the models was evaluated by the area under the ROC curve (AUC) and decision curve analysis (DCA). Results: A total of 282 patients were included, comprising 171 patients in the training set, and 111 patients in the validation set. Compared to the Sag-T2WI model (AUC, 95%CI, training set, 0.797, 0.722-0.782; validation set, 0.648, 0.521-0.776) and the Sag-T1C model (AUC, 95%CI, training set, 0.802, 0.723-0.882; validation set, 0.630, 0.505-0.756), the Ax-DWI model exhibited the highest diagnostic performance with AUCs of 0.855 (95%CI, 0.791-0.919) in training set, and 0.753 (95%CI, 0.638-0.867) in validation set, respectively. The combined model, integrating selected features from three sequences and FIGO stage, surpassed predictive ability compared to the single-sequence models, with AUC of 0.889 (95%CI, 0.833-0.945) and 0.859 (95%CI, 0.781-0.936) in the training and validation sets, respectively. Conclusions: The pretreatment MRI-based radiomics model, integrating radiomics features from three sequences and clinical variables, exhibited superior performance in predicting LNM following NACT in patients with LACC.
oncology