MIMOSA

Please run setup.m to start.

Pulseq Sequence

• Download Pulseq sequence programming environment (https://pulseq.github.io/) and add the matlab path.
• Run 01_gen_Pulse_Seq/3T/write_MIMOSA_1iso.m to generate Pulseq sequence file for 3T scan.
• Run 01_gen_Pulse_Seq/7T/write_MIMOSA_750um_iso_R4.m to generate Pulseq sequence file for 7T scan.

Reconstrcution

The baseline zero-shot reconstrcution code is forked from https://github.com/byaman14/ZS-SSL & https://github.com/yohan-jun/Zero-DeepSub

Installation

Dependencies are given in 02_Recon/3T/zsssl_recon_3T/environment_tf2.yml and can be installed withconda env create -f environment_tf2.yml.

Data

• The raw data of MIMOSA at R = 11.75 at 3T can be downloaded here. After downloading the raw data, put it in the folder 02_Recon/3T/rawdata.
• The raw data of MIMOSA at R = 4 at 7T can be downloaded here. After downloading the raw data, put it in the folder 02_Recon/7T/rawdata

Reconstrcution Pipeline

1. Preprocessing

• Run 02_Recon/3T/prepare_data_for_zsssl_recon.m and 02_Recon/7T/prepare_data_for_zsssl_recon.m to prepare data for 3T and 7T scans, respectively.

2. Training

• Run 02_Recon/3T/zsssl_recon_3T/zs_ssl_train_multi_mask_batch_v10_ms.py and 02_Recon/7T/zsssl_recon_7T/zs_ssl_train_multi_mask_batch_v10_ms.py to perform multi-contrast/-slice zero-shot self-supervised learning training for 3T and 7T scans, respectively. Prior to running training file, hyperparameters can be adjusted from parser_ops.py under the same path.

3. Inference

• Run 02_Recon/3T/zsssl_recon_3T/zs_ssl_inference_ms.ipynb and 02_Recon/7T/zsssl_recon_7T/zs_ssl_inference_ms.ipynb to load the check points saved during training for 3T and 7T scans, respectively.

Paramater Estimation

1. Run 03_ParamEstimation/3T/gen_MIMOSA_dict_3T.m and 03_ParamEstimation/7T/gen_MIMOSA_dict_7T.m to generate the dictionary for 3T and 7T scans, respectively.
2. Run 03_ParamEstimation/3T/MIMOSA_paramater_mapping_3T.m and 03_ParamEstimation/7T/MIMOSA_paramater_mapping_7T.m to perform paramater estimation process for 3T and 7T scans, respectively.

Copyright & License Notice

This project is licensed for non-commercial, research use only. For other purposes, please contact ychen156@mgh.harvard.edu.