MRI radiomics signature to predict lymph node metastasis after neoadjuvant chemoradiation therapy in locally advanced rectal cancer

摘要

Purpose To investigative the performance of MRI-radiomics analysis derived from T2WI and apparent diffusion coefficients (ADC) images before and after neoadjuvant chemoradiation therapy (nCRT) separately or simultaneously for predicting post-nCRT lymph node status in patients with locally advanced rectal cancer (LARC).Materials and Methods Eighty-three patients (training cohort, n = 57; validation cohort, n = 26) with LARC between June 2017 and December 2022 were retrospectively enrolled. All the radiomics features were extracted from volume of interest on T2WI and ADC images from baseline and post-nCRT MRI. Delta-radiomics features were defined as the difference between radiomics features before and after nCRT. Seven clinical-radiomics models were constructed by combining the most predictive radiomics signatures and clinical parameters selected from support vector machine. Receiver operating characteristic curve (ROC) was used to evaluate the performance of models. The optimum model-based LNM was applied to assess 5-years disease-free survival (DFS) using Kaplan-Meier analysis. The end point was clinical or radiological locoregional recurrence or distant metastasis during postoperative follow-up.Results Clinical-deltaADC radiomics combined model presented good performance for predicting post-CRT LNM in the training (AUC = 0.895,95CI:0.838-0.953) and validation cohort (AUC = 0.900,95CI:0.771-1.000). Clinical-deltaADC radiomics-postT2WI radiomics combined model also showed good performances (AUC = 0.913,95CI:0.838-0.953) in the training and (AUC = 0.912,95CI:0.771-1.000) validation cohort. As for subgroup analysis, clinical-deltaADC radiomics combined model showed good performance predicting LNM in ypT0-T2 (AUC = 0.827;95CI:0.649-1.000) and ypT3-T4 stage (AUC = 0.934;95CI:0.864-1.000). In ypT0-T2 stage, clinical-deltaADC radiomics combined model-based LNM could assess 5-years DFS (P = 0.030).Conclusion Clinical-deltaADC radiomics combined model could predict post-nCRT LNM, and this combined model-based LNM was associated with 5-years DFS in ypT0-T2 stage.