MICLab-2023-ISBI-MRS-Challenge

This repository contains the code for the Deep Spectral Divers team's first-place solution to the Edited-MRS Reconstruction Challenge presented at the IEEE International Symposium on Biomedical Imaging in 2023. It enables inference using the models used in the challenge, aiming to reconstruct GABA spectra faster by utilizing less data compared to current Edited-MRS scans. For each track, we trained a separate model, resulting in four final models: Track 01, Track 02, and two instances of Track 03, one with 2048 data points and the other with 4096 data points.

Our team was the overall winner of the challenge.

Challenge Description

Magnetic resonance spectroscopy (MRS) is a non-invasive method to quantify metabolite concentrations in vivo. It's particularly valuable for quantifying gamma-aminobutyric acid (GABA), a key inhibitory neurotransmitter in the brain. The detection of GABA is challenging due to interference from more abundant metabolites. Techniques such as spectral editing, spectral modeling, and two-dimensional MRS help isolate the GABA signal. Among these, MEGA-PRESS is a widely used technique for accurate GABA measurement.

Participants in the challenge were provided with simulated and in vivo data training sets representing GABA-edited MEGA-PRESS scans composed of two sub-signals (ON and OFF). Scripts for data augmentation, including adding noise, frequency, and phase shifts, were also provided. The models submitted by the teams were evaluated on simulated data (Track 01), homogeneous in vivo data (single-vendor) (Track 02), and heterogeneous in vivo data (multi-vendor) (Track 03) using quantitative metrics such as mean squared error, signal-to-noise ratio, linewidth, and peak shape.

The results of the challenge were presented at the IEEE International Symposium on Biomedical Imaging (ISBI) conference held in Cartagena, Colombia on April 18th, 2023. The challenge outcomes were summarized and submitted for a joint publication.

To access our training framework, please visit the GitHub repository for our Spectro-ViT model .

Our work on the Spectro-ViT model has been published in the Magnetic Resonance Imaging journal. You can find the full publication at this DOI link.

For more information about the Edited-MRS Reconstruction Challenge, visit the challenge webpage. Check the journal publication documenting the challenge results.

Table of Contents

• Installation
• Usage
• Citation

Installation

1. Clone the repository:

   git clone https://github.com/MICLab-Unicamp/MICLab-2023-ISBI-MRS-Challenge.git
   

2. Navigate to the project directory:

   cd MICLab-2023-ISBI-MRS-Challenge
   

3. Check the Python version in requirements.txt and install the required dependencies:

   pip install -r requirements.txt
   

Usage

Getting the predictions from the models:

1. Ensure you have the test data for each one of the tracks in the right .h5 format.

To execute the model's inference, you must have an h5 file containing the correctly formatted test data for each track (according to the challenge). The h5 file should be structured with the following datasets to ensure compatibility:

For Track 1:

• Dataset Name: "ppm"

  • Dataset Type: 1D array (np.ndarray)
  • Contains the ppm values.

• Dataset Name: "t"

  • Dataset Type: 1D array (np.ndarray)
  • Contains the time values.

• Dataset Name: "transients"

  • Dataset Type: 1D array (np.ndarray)
  • Contains the transients data.

For Track 2:

• Dataset Name: "ppm"

  • Dataset Type: 1D array (np.ndarray)
  • Contains the ppm values.

• Dataset Name: "t"

  • Dataset Type: 1D array (np.ndarray)
  • Contains the time values.

• Dataset Name: "transient_fids"

  • Dataset Type: 1D array (np.ndarray)
  • Contains the transients data.

For Track 3:

• Group Name: "data_2048"

  • Contains datasets for downsampled data.

  • Dataset Name: "ppm"

    • Dataset Type: 1D array (np.ndarray)
    • Contains the downsampled ppm values.

  • Dataset Name: "t"

    • Dataset Type: 1D array (np.ndarray)
    • Contains the downsampled time values.

  • Dataset Name: "transient_fids"

    • Dataset Type: 1D array (np.ndarray)
    • Contains the downsampled transients data.

• Group Name: "data_4096"

  • Contains datasets for upsampled data.

  • Dataset Name: "ppm"

    • Dataset Type: 1D array (np.ndarray)
    • Contains the upsampled ppm values.

  • Dataset Name: "t"

    • Dataset Type: 1D array (np.ndarray)
    • Contains the upsampled time values.

  • Dataset Name: "transient_fids"

    • Dataset Type: 1D array (np.ndarray)
    • Contains the upsampled transients data.

Make sure that the datasets are stored with the exact names and types mentioned above. The functions rely on these specific dataset names to extract the required data.

2. Obtain the trained weights by using our training framework.

3. Run the script:

Instructions for Tracks 01 and 02:

Execution:

python3 save_predicts_track01.py [weights] [test_data_path] [save_folder_path]

or

python3 save_predicts_track02.py [weights] [test_data_path] [save_folder_path]

Example usage:

python3 save_predicts_track02.py weights/weights_track02.pt data/track_02_test_data.h5 data/save_predicts

Replace [weights] with the path to the weights file for the track.

Replace [test_data_path] with the path to the track .h5 file containing the test dataset.

Replace [save_folder_path] with the folder path which the predict .h5 file will be saved.

Instructions for Track 03:

Execution:

python3 save_predicts_track03.py [weights_down] [weights_up] [test_data_path] [save_folder_path]

Example usage:

python3 save_predicts_track03.py weights/weights_track03_2048.pt weights/weights_track03_4096.pt data/track_03_test_data.h5 data/save_predicts

Replace [weights_down] with the path to the weights file for the track 03 downsampled (2048).

Replace [weights_up] with the path to the weights file for the track 03 upsampled (4096).

Replace [test_data_path] with the path to the track .h5 file containing the test dataset.

Replace [save_folder_path] with the folder path which the predict .h5 file will be saved.

4. The script will perform inference on each sample in the test dataset using the model.

5. The predicted spectra and ppm values will be saved in an output file named track01.h5, track02.h5, or track03.h5, depending on the respective script. These files will be located in the folder [save_folder_path] provided.

Developers

• Gabriel Dias
• Mateus Oliveira
• Lucas Ueda

Citation

Spectro-ViT model:

@article{DIAS2024,
title = {Spectro-ViT: A vision transformer model for GABA-edited MEGA-PRESS reconstruction using spectrograms},
journal = {Magnetic Resonance Imaging},
pages = {110219},
year = {2024},
doi = {https://doi.org/10.1016/j.mri.2024.110219},
author = {Gabriel Dias and Rodrigo Pommot Berto and Mateus Oliveira and Lucas Ueda and Sergio Dertkigil and Paula D.P. Costa and Amirmohammad Shamaei and Hanna Bugler and Roberto Souza and Ashley Harris and Leticia Rittner}
}

Edited-MRS Reconstruction Challenge:

@article{berto2023,
  title={Results of the 2023 ISBI challenge to reduce GABA-edited MRS acquisition time},
  journal={Magnetic Resonance Materials in Physics, Biology and Medicine},
  pages={1--15},
  year={2024},
  doi = {https://doi.org/10.1007/s10334-024-01156-9}
  author={Berto, Rodrigo Pommot and Bugler, Hanna and Dias, Gabriel and Oliveira, Mateus and Ueda, Lucas and Dertkigil, Sergio and Costa, Paula DP and Rittner, Leticia and Merkofer, Julian P and Van de Sande, Dennis MJ and others},
}