NeuroInfer

NeuroInfer (NI) is a tool for drawing Bayesian inferences on the associations between cognitive processes and brain activations based on meta-analytic data from fMRI published studies. In a nutshell, it allows neuroscientists to interpret their results and assess the reliability of their hypotheses. Compared to available tools (e.g., NeuroSynth, NeuroQuery, BACON), NI implements a more sophisticated Bayesian analysis (based on Bayesian confirmation theory). The NI framework includes a database of published fMRI studies and related metadata (updated in 2018 and shared with NeuroSynth and NeuroQuery), a vocabulary of terms, and a Python-based algorithm to perform the analysis.

GUI: A web-visualizer

Example of Uses:

First, be sure all the needed dependencies are correctly installed:

 $ pip install -r neuroinfer/requirements.txt

In order to open the gui on your own browser, just run the server.py script from your terminal:

 $ cd /your/path/to/neuroinfer/root
 $ pyhon -m neuroinfer.server

The python script analysis can be separatedely launched as follows:

 $ pyhon MainScript.py
 Enter cog(s) here (if more than one separate them with a comma): memory
 Enter a prior between 0.001 and 1.000 associated with memory: 0.1
 The analysis will be run on: memory with prior probability: 0.1.
 Do you want to insert coordinates (c) or the area (a)? a
 Choose an area from the list: 150
 Enter ROI radius here (in mm): 4
 Selected area: 150Radius: 4.0 mm.
 Specify the Bayesian confirmation measure you want to use in the analysis (type one of the following letter):
 'a' for Bayes' Factor;
 'b' for difference measure;
 'c' for ratio measure.
 a

Then the result is saved:

 $ Time in min: 0.03
 $ Results saved to: /home/saram/PhD/neuroinfer_brainhack23/neuroinfer/results/results_BHL_area150.pickle

To DO:

Analysis:

• general optimization of the analysis code
• implementation searchlight analysis
• implementation whole-brain analysis
• implementation of atlas-guided analysis

Results:

• back-end:

  • MainScript.py should be turned into init.py
  • UserInputs.py should be turned into a listener of the script form
  • index.html should be automatically open when the python script is run
  • create a python script to listen to the scripts output and lunch the anlaysis.

• front-end:

  • add a loading image in the web page while the analysis is run (and possibly estimate the time)
  • add an info box in the web page if the python script is detached
  • fix mni template showing when loading the page
  • implementing labeling overimposed to brain template (HTML5 & canvas)
  • implementing multi plots/results viewing/selection

Folder structure

./neuroinfer/
├── code/
│   │── plot_generator.py       # a template using flask_cors to interact with the javascript
│   │── DataLoading.py
│   │── BayesianAnalysis.py     # core bayesian anlaysis work
│   │── SummaryReport.py
│   │── UserInputs.py           # expects input from the command-line
│   └── MainScript.py           # main script (to be moved inside __init__.py)
├── html/
│   │── index.html              # this should be automatically open when the python 
│   │── main.js                 # provide the scripts to itneract with python
│   │── styles.css
│   └── viewer-config.js        # initialize neuroimaging div with mni template
├── templates/
│   └── {...}.nii.gz
├── data/                       # it contains the bibliography data needed to compute the analysis
│   │── featuures7.npz
│   │── metadata7.csv
│   │── vocabulary7.txt
│   │── data-neurosynth_version-7_metadata.tsv
│   └── data-neurosynth_version-7_coordinates.tsv
├── rii-mango-Papaya-#####/     # from papaya repo. it hosts the papaya visualizer js
├── images/     
└── README.md