v2.1.5: Minor colormap tweaks

Small tweaks to the lesion preview glassbrain colormaps, for consistency with newer versions of nilearn (in preparation for version update in the near future)

v2.1.4: Fix for sparse outputs, add experimental option for nonzero denom

Add experimental option "Only include subjects with non-zero denominator in local ratio". When computing the ratio for a given location or pair, only include subjects for which the denominator in that location (total number of streamlines ending there) is nonzero.

When enabled, it will affect all(chacovol|chacoconn)_(mean|stdev) outputs, but not any _allref outputs. It will also create an additional (chacovol|chacoconn)_denomfrac output, which is the fraction of subjects with a nonzero denominator for that location or pair. This can be applied offline by the user to threshold the chacovol_mean to exclude locations where only a small number of subjects have data. For instance, Vchaco[Vdenomfrac<0.01]=0 would set chaco values to zero for voxels where >=99% of the 420 reference subjects have no streamlines.

v2.1.3: Run both tractography algorithms, user-readable database location

• New option to run both probabilistic and deterministic tractography in the same upload
  • Results will include *_nemo_output_sdstream_*.* and *_nemo_output_ifod2act_*.* files
• Move database to new bucket to allow external downloads
  • New bucket is "request pays", so users can download the entire database (1.2 TB), using their own AWS account to cover data transfer fees, to allow running the tool on your own system.
  • Data is free to transfer to other AWS services within the same region (us-east-1)
  • Data transfer fees are charged (by Amazon, not by our group) to your billing account
    • Data transfer to another AWS region is $0.02/GB (approx $25 for 1.2TB)
    • Data transfer OUT of AWS is $0.09/GB for the first 10TB each month (approx $110 for 1.2TB)
    • See https://aws.amazon.com/s3/pricing/ for more details
  • You must have your own AWS account, with associated billing (no anonymous access)
  • Example command: aws s3 sync s3://kuceyeski-wcm-nemodata/ ./localnemo --request-payer requester
  • See https://docs.aws.amazon.com/AmazonS3/latest/userguide/ObjectsinRequesterPaysBuckets.html

v2.1.2: Add nemoSC_volnorm

Add nemoSC_volnorm with ROI-volume normalized estimated SC for specific parcellations (currently only fs86subj, cocommpsuit439subj, and shen268). In addition to *_nemoSC_mean.mat, outputs will include *_nemoSC_volnorm_mean.mat for those parcellations.

Also add "_sift2" to the nemoSC filenames when the "Weight streamlines by data fit (SIFT2)" option is selected, to more easily identify which estimated SC type was produced. Outputs will include *_nemoSC_sift2_mean.mat (and *_nemoSC_sift2_volnorm_mean.mat, as above).

NeMo v2.1.1

Freezing version v2.1a8 that has been stable. Changes since v2.1a7 include:

• Created a 91k subject-specific parcellation for the HCP MSMAll 32k_fs_LR space
• Had to modify website and main code to allow specifying ALL rois 1-91282 are filled, even if zeros
• now if atlas_list.json contains "regioncount" field, fill zeros for any missing in 1-regioncount
  (otherwise just include the unique non-zero values in input parcellation)
• also move accum+continuous options to "live" upload_internal page
• add cifti capability to chacovol_to_nifti.py
• Disable disable pairwise checkbox on website for res < 3mm (but allow override on upload_internal website)

v2.1.1 DOI: 10.5281/zenodo.11106578

NeMo v2.1

This version includes major bugfixes for both regional and pairwise outputs. The main effect of the bug was a downward shift in the scale of the regional Change in Connectivity (ChaCo) scores, which represent percent of disconnected streamlines per gray matter region. The magnitude of the shift in scale will depend on the atlas used, lesion mask patterns and the outputs requested when a user initially ran their lesion masks through the old version of the tool (pairwise, voxel-wise, regional etc). In our testing, new and old outputs were highly correlated, between 0.95 and 0.99 depending on lesion location, extent, and atlas selection, but may differ more in certain cases.

We also added new features for dilating atlases, which allows more streamline endpoints to be captured. The default behavior of many tractography tools is to expand gray-matter target ROIs to catch streamlines that end before reaching beyond white-matter. Previous NeMo outputs dilated gray-matter ROIs by 1mm. The new version allows users to choose between 0-3mm of dilation. We kept 1mm as the default for subject-specific gray matter parcellations (eg: fs86subj), and 0mm for MNI-defined parcellations (eg: AAL) for consistency with previous behavior, but users may find 2 or 3mm dilation better matches standard healthy subject connectome estimates.

In addition, when computing region-wise ChaCo scores, we now EXCLUDE all streamlines that start and end in the same region, as these are quite numerous and may not represent real underlying tracts. However, for a lobe-wise atlas or other atlas with very large regions (e.g. Yeo 7 or 17 functional networks), we recommend selecting the option to keep these streamlines. If users select both "pairwise" and "all reference subject" options upon lesion mask submission, they can re-compute region-wise outputs with or without these self-connections using the new "chacoconn_allref_to_chacovol.py" script.