Automated Apparent Diffusion Coefficient Measurement of Breast Tumors in Diffusion-weighted MRI for Early Neoadjuvant Chemotherapy Response Assessment

Purpose: To develop a fully automated method for measuring apparent diffusion coefficient (ADC) for breast tumors in diffusion-weighted MRI that is objective, repeatable, and reproducible for assessing early response to neoadjuvant chemotherapy. Materials and Methods: This study was a retrospective analysis of American College of Radiology Imaging Network 6698 trial data (August 2012–January 2015). Regions of interest (ROIs) were automatically derived by transferring the tumor ROI from dynamic contrast-enhanced MRI to diffusion-weight-ed MRI via image registration, and the ΔADC were calculated from baseline to early treatment. The ΔADC performance was assessed for predicting a pathologic complete response (pCR) using receiver operating characteristic curve analysis. The analysis was performed in all participants and in subgroups defined by tumor human epidermal growth factor receptor 2 (HER2) status. The reproducibility of automated tumor ADC measurements was assessed in a test-retest subcohort. Results: The analysis cohort included 226 participants with breast cancer (mean age, 48 years ± 10 [SD]). Automated ADC measurements were reproducible in the test-retest subcohort (n = 71; estimated agreement index, 0.82 [95% CI: 0.77, 0.85]). The ΔADC from the automated ROI predicted pCR, with an area under the receiver operating characteristic curve (AUC) of 0.61 (95% CI: 0.53, 0.69; P = .008). The AUC in the HER2-positive subcohort (0.69 [95% CI: 0.54, 0.93]) was higher than that in the HER2-negative subcohort (0.52 [95% CI: 0.41, 0.62]; P = .06). Conclusion: The automated ADC measurement method was reproducible, and tumor ΔADC could predict pCR early. External validation will be included in future work.