Comparison of Pathologic Response Evaluation Systems After Neoadjuvant Chemotherapy in Breast Cancers: Correlation With Computer-Aided Diagnosis of MRI Features.

Abstract

<b>OBJECTIVE.</b> The objective of our study was to compare pathologic response assessment systems with dynamic contrast-enhanced MRI (DCE-MRI) features and evaluate the predictive performance of DCE-MRI features relative to different pathologic response assessment systems after neoadjuvant chemotherapy (NAC). <b>MATERIALS AND METHODS.</b> This retrospective study included 223 women (mean age, 47.2 years; age range, 31-71 years) who underwent DCE-MRI before and after completing NAC before definitive surgery between January and December 2015. Six MRI features (i.e., tumor size; angio volume; peak enhancement; and persistent, plateau, and washout enhancing components) were measured and correlated with the Miller-Payne grading system, residual cancer burden (RCB) class, and modified in-breast RCB index. Percentage changes in MRI features were correlated with pathologic assessment systems using the Spearman rank correlation test; ROC curves were analyzed to predict pathologic outcomes. <b>RESULTS.</b> All six MRI features correlated most strongly with the in-breast RCB index (<i>r</i> = -0.75 to -0.51, <i>p</i> < 0.001) followed by the Miller-Payne system (<i>r</i> = 0.47-0.72, <i>p</i> < 0.001) and RCB class (<i>r</i> = -0.58 to -0.41, <i>p</i> < 0.001). The in-breast RCB index correlated most strongly with the angio volume reduction rate (<i>r</i> = -0.75, <i>p</i> < 0.001) followed by maximum diameter (<i>r</i> = -0.69, <i>p</i> < 0.001), peak enhancement (<i>r</i> = -0.67, <i>p</i> < 0.001), washout component (<i>r</i> = -0.60, <i>p</i> < 0.001), plateau component (<i>r</i> = -0.59, <i>p</i> < 0.001), and persistent component (<i>r</i> = -0.51, <i>p</i> < 0.001). <b>CONCLUSION.</b> The in-breast RCB index correlated best with changes in DCE-MRI features, and the MRI-measured angio volume reduction rate correlated best with pathologic tumor responses.