MAMA-Synth · Research demo
Predictive Enhancement Calibration
Source-only breast MRI virtual contrast enhancement with PEC and a reference-conditioned FLUX.2 latent generator.
Input. Upload a 2-D .mha / .npy / .npz, or a 3-D .nii / .nii.gz pre-contrast MRI. Auto mode accepts released MAMA-standardized arrays or applies the archived MAMA global z-score to raw intensities. For a 3-D volume, the app ranks slices by maximum intensity, samples from a Top-K pool, and lets you confirm or edit the selected indices before generation.
1 · Prepare the source MRI
Upload a file and confirm how its intensity values should be interpreted.
2 · Review and generate
For 3-D NIfTI, confirm up to four candidate slices before inference.
Upload a file or select one of the public examples below.
ZeroGPU uses the current visitor's daily quota. Signing in provides the free-account allowance; generation starts only when you click Generate.
Results
Run metadata
Generate an image to show the PEC coordinate and run details.
Public examples
Two deterministic examples from each MAMA-MIA collection, plus two Yunnan cases. Clicking a thumbnail only places its .mha in the input and prepares a preview—generation starts only when you click the button above.
Examples are modified pre-contrast slices from MAMA-MIA (overall CC BY-NC 4.0; source collection licenses apply) and the Yunnan cohort (CC BY 4.0). Exact selection, source IDs, licenses, and modifications are in manifest.json.
⚠️ Research use only. This is not a medical device and must not be used for clinical diagnosis, treatment, or decisions about contrast administration.