Initial experience in implementing quantitative DCE-MRI to predict breast cancer therapy response in a multi-center...

Area of Science:

• Medical Imaging
• Oncology
• Radiology

Background:

• Quantitative dynamic contrast-enhanced (DCE) MRI shows promise for predicting breast cancer response to neoadjuvant chemotherapy (NAC).
• Previous studies were primarily single-center and single-vendor, limiting generalizability.
• Implementing quantitative breast DCE-MRI in multi-center (MC) and multi-vendor platform (MP) settings is crucial for broader clinical application.

Purpose of the Study:

• To evaluate the feasibility and performance of quantitative breast DCE-MRI in multi-center, multi-vendor settings for predicting NAC response.
• To compare different pharmacokinetic (PK) analysis approaches, including tumor region of interest (ROI) vs. voxel-based analysis and the use of measured vs. fixed R1,0 values.
• To assess the predictive accuracy of various quantitative and semi-quantitative parameters against pathologic complete response (pCR).

Main Methods:

• Acquisition of B1 mapping, variable flip angle (VFA) R1 (R1,0) measurements, and high spatiotemporal resolution DCE-MRI during NAC across three sites with Siemens, Philips, and GE 3T platforms.
• Implementation of quality assurance/quality control (QA/QC) using a breast phantom.
• Pharmacokinetic (PK) analysis using the Tofts model (TM) and shutter-speed model (SSM), comparing tumor ROI- vs. voxel-based analyses and VFA-measured R1,0 vs. fixed R1,0.
• Evaluation of signal enhancement ratio (SER), longest diameter (LD), and PK parameters (Ktrans, kep) for predicting pCR.

Main Results:

• Voxel-based PK analysis using fixed R1,0 was identified as the optimal approach, accommodating data from a vendor with R1,0 overestimation issues.
• Semi-quantitative SER and quantitative PK parameters outperformed tumor longest diameter (LD) in predicting pCR after the first NAC cycle.
• Ktrans consistently showed higher prediction accuracy than SER and LD at both the first NAC cycle and midpoint.
• Both TM and SSM Ktrans and kep demonstrated excellent predictive performance at the NAC midpoint (AUC >0.90); SSM parameters showed better performance than TM after the first NAC cycle.

Conclusions:

• Quantitative breast DCE-MRI is feasible and effective in multi-center, multi-vendor settings for predicting NAC response.
• Voxel-based PK analysis with a fixed R1,0 value simplifies the process and mitigates cross-platform variability, potentially eliminating the need for B1 and VFA acquisitions.
• QA/QC with phantoms is essential for ensuring reliable MC/MP trial results.
• Pharmacokinetic parameters, particularly Ktrans and kep, are robust predictors of treatment response, with SSM showing advantages in early prediction.