Predicting pathological response after neoadjuvant chemotherapy of breast cancer using pharmacokinetic histogram features on dynamic contrast-enhanced magnetic resonance imaging

Abstract Background To investigate the ability of pharmacokinetic histogram features of dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) and clinical-pathological biomarkers for predicting pathological complete response (pCR) to NAC in breast cancer. Methods This retrospective study included 112 women with biopsy-proven breast malignancies from June 2019 to July 2020. The three-dimensional volume of interest tumors were drawn manually. A total of 51 pharmacokinetic histogram features were extracted and calculated for each participant from three pharmacokinetic parameters: K trans , K ep , and MaxSlope. Mann-Whitney U test, chi-squared test, logistic regression analyses, receiver operating characteristic (ROC) analysis, and Kendall's Tau-b correlation were performed. Clinical model, quantitative model, and combined model combining the pharmacokinetic histogram features and clinical-pathological factors for predicting pCR were constructed. The correlation between the clinical-pathological factors and independent risk pharmacokinetic histogram features was further observed. Results Simplified breast edema score (sBES), HER-2, Ki-67, K ep Maximum, and K trans Range were identified as independent predictors of pCR. The quantitative model comprised of K ep Maximum and K trans Range, as well as the combined model comprised of HER-2, K ep Maximum, and K trans Range, demonstrated good diagnostic performance and surpassed the clinical model comprised of HER-2 and sBES (0.880 vs 0.734, P = 0.009; 0.915 vs 0.734, P < 0.001). Additionally, significant correlations were observed between K ep Maximum and K trans Range and biologically aggressive clinical-pathological factors. Conclusions Using pharmacokinetic histogram features extracted from DCE-MRI improves the performance in predicting the pCR after NAC of breast cancer.