HierarchBayesParcel’s documentation

This repository implements the hierarchical Bayesian framework for learning brain parcellations across task-based and resting-state fMRI datasets. The technical details are described in the following paper, and we have applied the framework to generate a new probabilistic atlas of the human cerebellum.

The code for this framework is openly available. You can use this repository to:

• Learn new probabilistic brain parcellations across multiple fMRI datasets using other datasets for different brain structures.

• Use existing probabilistic atlases to obtain individualized brain parcellations for new subjects through the optimal integration of individual localizer data and the group atlas.

Diedrichsen Lab, Western University

Contents:

• Installation
  • Dependencies
  • Installation repository
• Overview
  • Arrangement Model
  • Emission Model
  • Full Model
  • Scope and related repositories
• Group atlas Training Example
  • Step 1: Define the the space in which to generate the group parcellations
  • Step 2: Prepare training data
  • Step 3: Build full model
  • Step 4: Fit the full model
  • Step 5: Visualize the group parcellations
  • Comparison of fusion atlas vs. single-dataset atlas
  • Monitoring the model training
  • Concentration parameter \(\kappa\)
  • Individual parcellations directly estimated from the group training
  • Avoid local maxima - use many optimization runs
  • GPU acceleration
• Individual Parcellation Example
  • Import all dependencies
  • Step 1: Define the the space in which to generate the individual parcellations
  • Step 2: Load the probabilistic group atlas
  • Step 3: Build an arrangement model
  • Step 4: Load individual localizing data
  • Step 5: Fit the emission model to the data
  • Step 6: Get the individual parcellation for a new subject
  • Step 7: Visualize the individual parcellations
• Mathematical details
  • Inference and learning
  • Arrangement models
  • Emission models
  • Model Evaluation
• GPU Acceleration
• API reference
  • Model module
  • Arrangement Module
  • Emission Module
  • Full_model module
  • Evaluation module
• References

Indices and tables

• Index

• Module Index

• Search Page

Repository

GitHub repository link: https://github.com/DiedrichsenLab/HierarchBayesParcel

License

Please find out our development license (MIT) in LICENSE file.

Bug reports

For any problems and questions, please use the issues page on this repository. https://github.com/DiedrichsenLab/HierarchBayesParcel/issues. We will endeavour to answer quickly.