Jan Valosek

> we "just" have to find what transformation was applied to the old vs. new PAM50, and apply it to the histology atlas. The PAM50 alignment to the ICBM152 was...

> Maybe the histology template was registered to the _previous_ version of the PAM50 This is right! The histology template is aligned to the [20180813_PAM50](https://osf.io/jmfpw): The PAM50 qfrom and sfrom...

I [computed the qform and sform difference](https://github.com/neuropoly/template/blob/jv/align_histological_template_to_ICBM152_PAM50/update_header_PAM50.py) between the [20180813_PAM50](https://osf.io/jmfpw) and the [current ICBM152 PAM50](https://github.com/spinalcordtoolbox/PAM50/tree/master/template) and applied it to the histological template. And it seems to be working: Now, the...

> I confirm coverage is C1-C6 Okay! Thank you! I updated the header for all histological template files and went back to the https://github.com/spinalcordtoolbox/spinalcordtoolbox/issues/2179#issuecomment-1387530873: > thank you for looking into...

> > The last question; do we want to threshold the histological template? E.g., by 0.01 threshold? To remove interpolation errors? (note: these interpolation errors were already presented also in...

> Please truncate the NIfTI files to headers. > > ``` > zcat img.nii.gz | head -c 352 | gzip - | sponge img.nii.gz > ``` Thanks! Done in https://github.com/bids-standard/bids-examples/pull/405/commits/6a84c8f50d30bfd997095859d852f3c305972707

> Could you update the dataset listing? > > https://github.com/bids-standard/bids-examples/blob/master/CONTRIBUTING.md#how-to-generate-the-tables-in-the-readme > > It is in the root of the repo: https://github.com/bids-standard/bids-examples/blob/master/dataset_listing.tsv Done in https://github.com/bids-standard/bids-examples/pull/405/commits/ec44bc3a6c182b26f92b601fe133e89ca1a44d40. > Ideally it should link to...

Since the [corresponding PR](https://github.com/bids-standard/bids-specification/pull/1586) has been merged, maybe this PR could be merged as well?

Just a follow-up: when trying to run the trained nnU-Net model directly using pytorch using `torch.device("mps")`, I get `ConvTranspose 3D is not supported on MPS`. ```python File "/Users/valosek/miniconda3/envs/nnunet/lib/python3.9/site-packages/dynamic_network_architectures/building_blocks/unet_decoder.py", line 109,...

We came across this point again with @tzebre and @jcohenadad when discussing if segmentations produced by the nnU-Net framework should be accompanied by JSON sidecar (see discussion [here](https://github.com/ivadomed/utilities/issues/20#issuecomment-1610293930)). Since this...