Development and validation of a radiomics model based on lymph-node regression grading after neoadjuvant chemoradiotherapy in locally advanced rectal cancer.

SiYu Zhang,Bin Tang,MingRong Yu,Lei He,Ping Zheng,ChuanJun Yan,Jie Li,Qian Peng

DOI: https://doi.org/10.1016/j.ijrobp.2023.05.027

2023-05-25

Abstract:Purpose The response to neoadjuvant chemoradiotherapy (nCRT) varies among locally advanced rectal cancer(LARC) patients, and the treatment response of lymph nodes(LNs) to nCRT is critical in implementing a watch-and-wait(WW) strategy. A robust predictive model may help personalize treatment plans to increase the chance patients achieve a complete response. This study investigated whether radiomics features based on pre-nCRT MRI nodes could predict treatment response in preoperative LARC LNs. Patients and Methods The study included 78 patients with clinical stage T3-T4, N1-2, and M0 rectal adenocarcinoma who received long-course neoadjuvant radiotherapy before surgery. Pathologists evaluated 243 LNs, of which 173 and 70 were assigned to training and validation cohorts, respectively. For each LN, 3641 radiomics features were extracted from the region of interest in high-resolution T2WI magnetic resonance imaging (MRI) prior to nCRT. The least absolute shrinkage and selection operator (LASSO) regression model was used for feature selection and radiomics signature building. A prediction model based on multivariate logistic analysis, combining radiomics signature and selected LN morphological characteristics, was developed and visualized by drawing a nomogram. The model's performance was assessed by receiver operating characteristic (ROC) curve analysis and calibration curves. Results The radiomics signature consists of 5 selected features that were effectively discriminated within the training cohort (AUC 0.908, 95%CI 0.857-0.958) and the validation cohort (AUC 0.865, 95%CI 0.757-0.973). The nomogram, which consisted of radiomics signature and LN morphological characteristics (short-axis diameter and border contours), showed better calibration and discrimination in the training and validation cohorts (AUC 0.925, 95% CI:0.880-0.969 and AUC 0.918, 95% CI:0.854-0.983, respectively). The decision curve analysis confirmed that the nomogram had the highest clinical utility. Conclusion The nodal-based radiomics model effectively predicts LNs treatment response in LARC patients after nCRT, which could help personalize treatment plans and guide the implementation of the WW approach in LARC patients.

oncology,radiology, nuclear medicine & medical imaging