kwyk2nidm

NIDMification of kwyk output

This project uses kwyk (https://github.com/neuronets/kwyk). Paper, code, and model corresponding to preprint, which is now published.

Cite: McClure P, Rho N, Lee JA, Kaczmarzyk JR, Zheng CY, Ghosh SS, Nielson DM, Thomas AG, Bandettini P and Pereira F (2019) Knowing What You Know in Brain Segmentation Using Bayesian Deep Neural Networks. Front. Neuroinform. 13:67. doi:10.3389/fninf.2019.00067

Steps

Run kwyk

Interpret kwyk output for regional volumes

We include a BASH script, 'kwykput.sh' that takes a resulting output from kwyk, and used the FSL fslstats utility to determine the volume for each of the regions. This script uses the kwyk_region_list.txt file for the region lables (derived from FreeSurfer. It generates a text file (example provided test_out.txt) of the form:

kwyk_index label number_voxels vol_inmm3
1 Cerebral-White-Matter 496396 496396.000000 
2 Ventricular-System 11025 11025.000000 
3 Cerebellum-White-Matter 32515 32515.000000 
4 Cerebellum-Cortex 144992 144992.000000 
5 Thalamus-Proper 18118 18118.000000 
6 Caudate 10851 10851.000000 
etc...

Convert the volume result file into NIDM

The steps for this include generating a kwykmap.json file that described the content of out reults file (kwykmap.json). As a developer, you can then work with ReproNim to create a set of custom terms for your software, which are harmonized with other software. This generates a list of custom data elements for the software with additional information. For example, in the case of kwyk, this maps every structure to a common anatomical name, measurement type, and measurement units.

kwyk:kwyk_000002 a kwyk:DataElement ;
    kwyk:label "Cerebral-White-Matter vol_inmm3 (mm^3)" ;
    kwyk:measure "vol_inmm3" ;
    kwyk:structure "Cerebral-White-Matter" ;
    kwyk:structure_id 1 ;
    kwyk:unit "mm^3" ;
    nidm:datumType ilx:0738276 ;
    nidm:hasUnit "mm^3" ;
    nidm:isAbout uberon:0002437 ;
    nidm:measureOf ilx:0112559 .

Using the kwyk data elements to generate a NIDM result

Software environment

You can run the kwyk2nidm script using either of the methods below.

1. Install kwyk2nidm into your Python 3 environment

pip install https://github.com/ReproNim/kwyk2nidm/archive/master.zip

kwyk2nidm -f kwyk_stats_file

2. Clone the repo and create a Docker container

git clone https://github.com/ReproNim/kwyk2nidm.git
cd kwyk2nidm
docker build -t kwyk2nidm:latest .
docker run -v $(pwd):/data kwyk2nidm -f /data/kwyk_stats_file

Running the kwyk2nidm command will generate an output file kwyk_stats_file.ttl unless a different name is specified using the -o flag.

To generate all the NIDM KWYK data elements add -g to the commands above. This will generate a KWYK-NIDM.ttl that should be added to triplestore to provide the link between the NIDM stats file to the common data attributes.

Great! I have a NIDM kwyk result. Now What???

1. Upload to ReproPond or ReproLake
2. Query across files!
3. Merge it with other NIDM, and query it!