Abstract

Motivation: Determine best-practice quantitative DCE-MRI for predicting breast cancer (BC) response to neoadjuvant chemotherapy (NAC) in a multi-center (MC) and multi-vendor platform (MP) setting. Goal(s): Evaluate effects of different pharmacokinetic analysis approaches on Ktrans and its predictive performance. Approach: 15 BC patients treated with NAC underwent longitudinal DCE-MRI at 3 sites using 3T systems from 3 vendors. Variations in analysis included Tofts model vs. Shutter-Speed model (SSM), ROI- vs. voxel-based analysis, and using fixed vs. measured R10. Results: Different analysis approaches resulted in significantly different Ktrans, with SSM Ktrans from voxel-based analysis using fixed R10 showing highest predictive accuracy for response. Impact: Voxel-based SSM analysis using fixed R10 takes advantage of greater range of SSM Ktrans changes in response to therapy, mitigates R10 measurement errors, and may be the best-practice quantitative DCE-MRI for predicting NAC response in a MC and MP setting.