Radiomic Based On Dynamic Contrast Enhanced MRI For Early Prediction Of Axillary Response Of Breast Cancer After Neoadjuvant Chemotherapy

Objective: The objective of this paper is to explore the value of radiomic models using dynamic contrast-enhanced(DCE)MRI for early prediction of the axillary lymph node(ALN)pathological complete response(p CR)of breast cancer patients after neoadjuvant chemotherapy(NAC).Methods: A total of 224 patients with ALN-positive breast cancer who underwent breast MRI before and one cycle after NAC treatment from October 2018 to January 2022 were prospectively included.Patients were divided into a training(n = 156)and validation(n =68)cohort based on the temporal order of their treatments.The radiomics of the primary tumor,transition zone and lymph node were extracted from DCE-MRI images before(pre-)and after the first cycle(1st-),respectively,and their change(delta-)radiomic values were calculated and recorded.A logistic regression model was used to build the pre-,1st-,and delta-radiomic based on the primary tumor,transition zone and lymph node,respectively,and their fusion radiomic models.The area under the receiver operating characteristic ROC curve(AUC)was used to assess the predictive performance of the different radiomic models,and the De Long test was used to compare the differences between the models.Results: In all radiomic models,the delta-radiomic model showed good prediction performance in primary tumor,transition zones and lymph node and three regions of interest(ROI)fusions both in the training and validation cohorts.Based on the deltaradiomic model obtained the highest prediction performance compared to the pre-,1stradiomic model.The predictive performance of lymph node in the training cohort was improved in all radiomic models when the ROI of primary tumor + transition zone was fused(Delong test: all p<0.05).The delta-radiomic model with primary tumor + transition zone + lymph node achieved the best predictive performance when combined with clinical features,with AUC values of 0.873(95% CI: 0.818-0.927)and 0.834(95% CI: 0.738-0.930)in the training and validation cohort,respectively.The calibration curve and decision curve also showed good agreement and high clinical benefit of the combined model.Conclusion: This study shows that delta-radiomic model based on primary tumor +transition zone + lymph node combined with clinical features can effectively predict axillary p CR after the first cycle of NAC in patients with breast cancer.