0704

The first MR Electrical Properties Tomography (MR-EPT) reconstruction challenge  
Stefano Mandija1,2 and Cornelis A.T. van den Berg1,2
1Department of Radiotherapy, UMC Utrecht, Utrecht, Netherlands, 2Computational Imaging Group for MR Diagnostics and Therapy, UMC Utrecht, Utrecht, Netherlands

Synopsis

This work announces the initial design of the first Electrical Properties (EPs) reconstruction challenge and opens up the call for ideas to finalize its implementation and to define the organizing committee members. Several EPs reconstruction methods have been presented in the last decade. However, as emerged from the last EPs workshop (IMEP-Utrecht’19), testing on common data has not been performed yet, nor data are publicly available. The scope of this challenge is therefore to benchmark current reconstructions on the same data. These data will also be made available to favor benchmarking of new methods in the future.

Introduction

MR-Electrical Properties Tomography (MR-EPT) aims at reconstructing tissue electrical properties (EPs), (mostly) conductivity and permittivity, at megahertz frequencies from non-invasive MRI measurements. In the last decade, several reconstruction methods have been presented, however the reported EPs values show substantial differences1,2,3. This was recognized during the last EPT workshop (IMEP-Utrecht-2019)4, demonstrating the need of a common dataset to benchmark current reconstruction algorithms and future physics/deep-learning based reconstruction methods. For this purpose, the Utrecht group set out to provide such dataset (simulated and measured phantom/in-vivo data) fulfilling the community-shared key requirements: use of a clinical MRI system (3 T), standard MRI sequences and clinical coil setup to ease future clinical implementation.Additionally, during that workshop it was recognized the need to objectively compare the current reconstruction methods to understand their strengths and weaknesses in a systematic manner. In consultation with the future EMTP workshop organizers (Lucca-2022), it seems that launching the first MR-EPT reconstruction challenge in the context of this workshop would help reaching these overarching goals. Here, we therefore present the initial design of the first MR-EPT reconstruction challenge. We also invite the EPT community to provide suggestions to be discussed during the upcoming ISMRM Meeting (London-2022) through the call for ideas and to candidate for the organizing committee (see below).

Methods

Challenge Phases: Following the lessons learned in the QSM reconstruction challenges5,6, the EPT reconstruction challenge will comprise of three independent phases (see Fig. 1 for description, data provided, and evaluation strategy of each phase).
Timeline: We foresee that the challenge can be launched in the context of the EMTP workshop (Lucca-2022) and results can be presented in the following ISMRM Meeting.
Dataset: As for the QSM challenges, we aim at providing different type of data for brain EPs reconstructions at 3 T:
1) In-silico brain data (noiseless and with noise) from Sim4Life simulations:
  • B1+ magnitude, phase and receive phase (for transceive phase-based reconstructions), source fields
  • Ground-truth (GT) EPs maps
2) MRI cylindrical phantom data:
  • B1+ magnitude and transceive phase
  • GT-EPs maps
  • MRI magnitude data (e.g. T1-weighted images)
3) MRI 3D-printed brain-like phantom data:
  • B1+ magnitude and transceive phase
  • GT-EPs maps
  • MRI magnitude data (e.g. T1-weighted images)
4) MRI in-vivo brain data (3 healthy volunteers):

  • B1+ magnitude and transceive phase
  • MRI magnitude data (e.g. T1-weighted images)
Data Management: As for the QSM challenge 2.0, data evaluation will be implemented to allow a fully blinded analysis. Only two committee members will have access to the identifying information of the participants. Upon submission of the reconstruction results, participants will also be asked to answer a questionnaire similar to the one presented in the QSM challenge 2.0 providing further information on the type of reconstruction performed and used data (e.g. phase-only).
Metrics: As learned in the QSM challenges, we will not only implement quantitative, global error metrics (e.g. NRMSE). In addition, region specific analysis and visual rating will also be considered. Such rating will provide a quality assessment of the reconstructed EPs maps, lack of reconstruction artifacts (boundary errors / blurring), presence of wrong morphological structures.
Open Science: We also strive for open science. Therefore, we will aim at making our database publicly available (downloadable after registration and user verification). Also, analogously to the QSM challenge 2.0, we will encourage all the participating groups to make available the codes and/or to report about their willingness to provide them upon request.

Results

Quantitative and qualitative analyses (Fig.1) will be performed by the committee members and results will ideally be presented at the ISMRM Annual Meeting 2023. Decisions on the final design and metrics will be taken by the organizing committee after reviewing the input from the EPT community (Fig.2).

Discussion

Through this challenge, we expect to achieve the following major goals:
1) Provide the first dataset for MR-EPT reconstructions fulfilling the main requirement of: using a clinical MRI system (3 T), standard MRI sequences and clinical coil setup to ease future clinical implementation;
2) Evaluate the state-of-the-art of EPT reconstructions methods using the same test data;
3) Allow benchmarking of future EPT reconstruction methods.
At the current stage, we are looking forward to hearing possible suggestions from the EPT community (e.g. MRI sequences to be used) as well as understanding the overall interest in participating in such a challenge. For this purpose, we invite the readers interested in participating to such a challenge to fill in the following survey (Google form: https://forms.gle/XJV2C65b8wrUGpUg7 – call for ideas, Fig.2). We aim at discussing the collected ideas during the upcoming ISMRM Annual Meeting (London-2022). Also, if you would like to be part of the organizing committee, we kindly invite you to candidate via the same form.

Conclusion

This work presents the initial design of the first MR-EPT reconstruction challenge and opens up the call for ideas to finalize its implementation and to define the organizing committee members.

Acknowledgements

The authors thank Dr. Zilberti Luca for the constructive discussion in setting up the challenge idea and its hosting during the next EMTP workshop (Lucca-2022). This work was partially supported by the Netherlands Organisation for Scientific Research (NWO), grant number:18078.

References

1) Hancu I, Liu J, Hua Y, Lee SK, ‘Electrical Properties Tomography: Available contrast and reconstruction capabilities.’ Magn Reson Med 2019; 81: 803-810.

2) McCann H, Pisano G, Beltrachini L, ‘Variation in reported human head tissue electrical conductivity values.’ Brain Topography 2019; 32: 825-858 (original article)

3) Mandija S, Petrov PI, Vink JJT, Neggers SFW, van den Berg CAT, ‘Brain tissue conductivity measurements with MR-Electrical Properties Tomography: An in vivo study.’ Brain Topography 2021; 34: 56-63

4) Event recorded in the EMTP community website: EMTP Hub - https://www.emtphub.org/events/

5) Milovic C, Tejos C, et al, ‘The 2016 QSM Challenge: Lessons learned and considerations for a future challenge design.’ Magn Res Med 2020; 84: 1624-1637

6) QSM Challenge 2.0 Organizing Committee, et al, ‘QSM reconstruction challenge 2.0: Design and report of results.’ Magn Res Med 2021; 86: 1241-1255

Figures

Fig.1: Description of the 3 phases of the MR-EPT challenge. GT-EPs = ground-truth electrical properties.

Fig.2: List of questions present in the Google form to collect suggestions and to candidate for the organizing committee.

Proc. Intl. Soc. Mag. Reson. Med. 30 (2022)
0704
DOI: https://doi.org/10.58530/2022/0704