YODA / Regression is all you need for medical image translation

Model weights and singularity/apptainer environment for YODA (Regression is all you… and extensions (ISMRM 2026).

Please note the usage instruction at github.com/Deep-MI/YODA. The provided dagobah.sif file is a pre-build Singularity/Apptainer container, ie the result from singularity build dagobah.sif docker://srassmann/dif:latest.

If you use the resources in your research, please always cite the YODA paper + checkpoint-specific additional (conference) papers.

Checkpoint	Trans. Task	Resolution	Train. Dataset	Train. Paradigm	Citation(s)
rs_FLAIR_from_T1T2.zip	T1w+T2w -> FLAIR	1 mm	RS (n=1344)	Diffusion	YODA
brats_FLAIR_from_T1T2.zip	T1w+T2w -> FLAIR	1 mm (resampled)	BraTS '23 (n=1270)	Diffusion	YODA
rs_FLAIR_from_T1.zip	T1w -> FLAIR	1 mm	RS (n=1344)	Diffusion	YODA, ISMRM 2026 (FLAIR)
rs_FLAIR_from_T2.zip	T2w -> FLAIR	1 mm	RS (n=1344)	Diffusion	YODA, ISMRM 2026 (FLAIR)
ixi_T2_from_T1PD.zip	T1w+PD -> T2w	~1 mm	IXI (n=511)	Diffusion	YODA
GoldAtlas_CT_from_MR.zip	T1w+T2w -> CT	~1x1x3 mm	Gold Atlas (n=11)	Diffusion	YODA
rs_T1_from_FLAIR.zip	FLAIR -> T1w	1 mm	RS (n=2500)	Regression	YODA, ISMRM 2026 (T1w)
rs_T1_from_T2w.zip	FLAIR -> T2w	1 mm	RS (n=2500)	Regression	YODA, ISMRM 2026 (T1w)

 

Citations:

YODA: Rassmann et al. (2026) "Regression is all you need for medical image translation", IEEE Transactions on Medical Imaging

ISMRM 2026 (FLAIR): Rassmann et al. (2026) "FLAIR-less white-matter hyperintensity segmentation using YODA", ISMRM 2026 (Cape Town)

ISMRM 2026 (T1w): Rassmann et al. (2026) "MRI contrast translation for full-brain segmentation from T2-weighted contrasts", ISMRM 2026 (Cape Town)

 

Note: Depending on the Training Paradigm, the checkpoints require either the diffusion (dm_predict.py) or regression (reg_predict.py) inference scripts.

 

The file expected_output-FLAIR_from_T1T2.nii.gz is obtained from running the respective T1w+T2w->FLAIR translator on the example RS case.