Welcome to Localise’s documentation!

LOCALISE is a Python library for segmentating prevalent surgical targets using multimodal MRI. This toolbox uses Image Quality Transfer techniques to transfer anatomical information from high-quality data to a wide range of connectivity features in low-quality data. The goal is to augment the inference on DBS targets localisation even with compromised data quality.

At present, the toolbox supports:

• Localising the Vim nucleus of the thalamus, given the T1 and the diffusion MRI data.

• Training a model for localising a new target, given the T1 and the diffusion MRI data.

Future releases will include:

• Localising the STN, ACC, and other targets…

Note

This project is under active development, and the API might still change substantially at any time!

How to Get Started

1. If the toolbox is not installed, following the Installation Instructions.

2. If you are not sure what data you need and how to prepare them, check out the Prepare Your Data.

3. If you simply want to localise a target using pre-trained models, check out the Predict Mode. Look through the tutorials.

4. If the toolbox doesn’t include the target you want to localise, and you want to train a custom model tailored for your own data, check out the Train Mode.

5. For more information about the toolbox, check out our paper.

Contents:

• Installation
  • FSL and Freesurfer Installation
• Usage
  • Introduction
  • Prepare Your Data
    • Prerequisites
    • Estimate fiber orientations
    • Create anatomical masks
    • Generate tract-density maps
  • Localise a Target (Predict Mode)
    • Prerequisites for Data Organisation
    • Tutorials for Localising a Target
  • Train a Custom Model (Train Mode)
    • Prerequisites for Custom Model Training
    • Tutorials for Training a Custom Model
• API
  • localise
• References

Contributors

The original toolbox was written by Ying-Qiu Zheng.

Other Contributors:

• Saad Jbabdi