The 2016 QSM Challenge: Lessons learned and considerations for a future challenge design

Area of Science:

• Medical Imaging
• Biophysics

Background:

• The 4th International Workshop on MRI Phase Contrast and QSM hosted the inaugural QSM Challenge.
• A gradient recalled echo acquisition dataset was provided with COSMOS and χ33 STI as ground truths.

Purpose of the Study:

• To evaluate submitted Quantitative Susceptibility Mapping (QSM) solutions from the first QSM Challenge.
• To investigate the impact of background field remnants and anisotropic susceptibility contributions on QSM reconstructions.
• To assess the influence of noise on metric scores using forward simulations.

Main Methods:

• Applied additional filters to investigate background field remnants.
• Estimated anisotropic contributions to amend the χ33 ground truth.
• Utilized forward simulations from COSMOS reconstruction to analyze noise impact on metric scores.

Main Results:

• Reconstructions showed lower errors when compared against the amended STI ground truth.
• Background field remnants had a minimal effect on reconstruction errors.
• Structural Similarity Index Metric demonstrated greater robustness to inconsistencies.

Conclusions:

• A discrepancy exists between provided data and ground truths due to anisotropic susceptibility and noise.
• Simulations are proposed for future QSM Challenges owing to the unreliability of in vivo ground truths.