Search papers, labs, and topics across Lattice.
This study introduces Predictive Enhancement Calibration (PEC) to improve virtual contrast enhancement (VCE) of breast MRI by addressing the limitations of existing latent generators, which struggle with the non-canonical intensity scale of MRI. By training a shared, case-adaptive coordinate system and predicting unavailable upper endpoints from pre-contrast images, PEC enhances radiomic fidelity during image synthesis. The method outperforms traditional approaches in a source-only VCE setting, demonstrating significant improvements in key metrics such as MSE and LPIPS on the MAMA100 development cohort.
PEC boosts the fidelity of virtual contrast-enhanced breast MRIs by effectively recalibrating intensity scales, outperforming conventional methods in critical evaluation metrics.
Virtual contrast enhancement (VCE) synthesizes enhanced breast MR images from pre-contrast acquisitions. Modern latent generators offer strong image priors, but their bounded natural-image autoencoders conflict with the non-canonical intensity scale of MRI. We show that the upper bound can alter radiomic fidelity before generation, while scaling source and target independently creates a coordinate inconsistency. We propose Predictive Enhancement Calibration (PEC), which represents each pair in a shared, case-adaptive coordinate during training and predicts its unavailable upper endpoint from the pre-contrast image at inference. We integrate PEC with a pretrained FLUX latent flow transformer via parameter-efficient reference conditioning. Target round trips first isolate representation loss before generation; near-matched conditional models then compare PEC with fixed-wide and separate coordinates under comparable training budgets and backbone settings. On the fixed internal MAMA100 development cohort, PEC improves all eight point estimates in this source-only VCE setting, with paired evidence strongest for MSE and LPIPS.\noindent\textbf{Code:} https://github.com/tanlei0/pec-breast-mri-vce
