Retrospective Correction of ADC for Gradient Nonlinearity Errors in Multicenter Breast DWI Trials: ACRIN6698...

Area of Science:

• Medical Imaging
• Quantitative MRI
• Oncology

Background:

• Diffusion-weighted magnetic resonance imaging (DWI) provides quantitative metrics like the apparent diffusion coefficient (ADC).
• Interplatform variability in ADC measurements, caused by systematic gradient nonlinearity (GNL), limits its clinical utility in multi-site trials.
• Accurate ADC quantification is crucial for predicting breast cancer treatment response.

Purpose of the Study:

• To assess the feasibility and effectiveness of a retrospective GNL correction (GNC) method.
• To reduce interplatform variability in ADC measurements from DWI.
• To improve the quantitative accuracy and reproducibility of ADC metrics for multiplatform clinical oncology trials.

Main Methods:

• A retrospective GNL correction (GNC) was implemented using system-specific gradient-channel fields.
• The GNC method was applied to quantitative quality control phantom data and a subset of 60 subjects from the ACRIN 6698 breast cancer trial.
• Trace-DWI data were analyzed using vendor-provided spherical harmonic tables for GNL correction.

Main Results:

• GNC significantly improved interplatform accuracy of ADC measurements in phantoms from 6% to 0.5%.
• Reproducibility of ADC measurements in phantoms improved from 11% to 2.5% after GNC.
• In trial subjects, GNC increased low ADC tumor volume by 16% and shifted ADC thresholds by ~0.06 µm²/ms.

Conclusions:

• Retrospective GNL correction is feasible and effective for improving ADC accuracy and reproducibility.
• GNC enhances the quantitative utility of ADC metrics in multiplatform clinical imaging trials.
• This approach supports more reliable breast cancer treatment response prediction using quantitative DWI.