B2RIO: "Brain t(w)o Reverse Inference Ontology" package

How to use this package

Instalation

Clone this repository and install the application with python setup.py install

Running the analysis

The only necessary parameters are the path to a nifti file with the image to be analyzed and the name of the output file. The program expects a 4D image with the region's labels (or probabilities in the probabilistic version) in the voxels we want to evaluate and zero in all the other voxels.

E.g: b2rio --brain_path nifti_image.nii.gz --output_file results.

Apart from this, there are a number of other parameters that can be configured:

• --n_folds: Number of folds use to sample studies (default=150)
• --frac_sample: Percentage of studies to use in folds (default=0.7)
• --resample: Resample of the provided image (default=1)
• --radius: When matching voxels between the provided image and neurosynth activations, this radius is used to limit the maximum distance between co-activated voxels. (default=4)
• --tfIdf: Threshold to limit the number of studies based on the tfIdf provided by Neurosynth.
• --output_summary: If true, an {output_file}_summary.csv file will be calculated containing the mean, std and skew between the folds of each term.

Examples

Analizing the Area STS1 (STS) of the Julich's atlas.

Deterministic atlas

Suppose we want to obtain the cognitive processes associated with the STS1 region of the left hemisphere of the Julich atlas. To do this, we must first obtain the corresponding image. It can be obtained directly from the official Julich atlas repository, with the following command:

curl -O https://object.cscs.ch/v1/AUTH_227176556f3c4bb38df9feea4b91200c/hbp-d000001_jubrain-cytoatlas-Area-STS1_pub/5.3/Area-STS1_l_N10_nlin2ICBM152asym2009c_5.3_publicDOI_a8128c417aeeb34c81dabc225670faa5.nii.gz

Then, it only remains to run B2RIO with the name of the nifti file as the --brain_path parameter and a name for the output file using the --output_file parameter.

b2rio --brain_path Area-STS1_l_N10_nlin2ICBM152asym2009c_5.3_publicDOI_a8128c417aeeb34c81dabc225670faa5.nii.gz --output_file STS1

When the process is complete, it will return a CSV file named STS1.csv with our results. The following are the first ten entries of the obtained file (out of a total of 18600, remember that the number of folds can be modified with the --n_folds parameter)

	term	fold	bf	topConcept
0	action	0	0.937732	Action
1	action	1	0.83446	Action
2	action	2	0.870331	Action
3	action	3	0.834874	Action
4	action	4	0.852323	Action
5	action	5	0.86945	Action
6	action	6	0.920112	Action
7	action	7	0.82419	Action
8	action	8	0.852273	Action
9	action	9	0.885683	Action

Moreover, if we organize the results by their Bayes Factor in a decreasing order, we can see that the most relevant results are related to cognitive processes associated with listening/audition (confirming in some way the knowledge available in the literature).

	term	fold	bf	topConcept
9256	listening	106	2.5791
9192	listening	42	2.56588
9164	listening	14	2.53278
9269	listening	119	2.52988
9178	listening	28	2.5297
8984	language comprehension	134	2.51597	Language
9280	listening	130	2.51489
9265	listening	115	2.50568
8929	language comprehension	79	2.50209	Language
9167	listening	17	2.49724

Probabilistic atlas

If our atlas has probabilities associated to its voxels and we want to perform the same analysis but taking into account this information, we only need to run B2RIO in its probabilistic version using the command b2rio_prob.

b2rio_prob --brain_path Area-STS1_l_N10_nlin2ICBM152asym2009c_5.3_publicDOI_a8128c417aeeb34c81dabc225670faa5.nii.gz --output_file STS1_prob

As an example, if we organize the results as before, these are the ten entries with the highest Bayes Factor:

	term	fold	bf	topConcept
8929	language comprehension	79	5.34525	Language
9256	listening	106	5.26801
9164	listening	14	5.25006
8868	language comprehension	18	5.1464	Language
8978	language comprehension	128	5.07841	Language
15374	sentence comprehension	74	5.07644	Language
8901	language comprehension	51	4.9826	Language
8984	language comprehension	134	4.98044	Language
9192	listening	42	4.96508
9265	listening	115	4.95206

Issues/Improvements

If you find a bug or think that B2RIO can be improved in any way, feel free to open an Issue or send a PR