[DATA SUBMISSION] Fitting challenge data

Fitting challenge data.Synthetic data mimicking single-voxel 3 T PRESS TE = 30 ms MRS spectra.


Dear @gosia,

Thank you so much - this is an outstanding submission, thank you so much for making the Fitting Challenge data available. I hope everyone is going to benchmark their software with these data! :slight_smile:

I have added an entry for your submission to the MRSHub Dataset Collection.

@gosia I personally think that synthetic data are a really useful way to explore the modelling performance of the various fitting tools that we have at our hands, particularly now that we have a bunch of new ones emerging. Are you planning on developing the Fitting Challenge? There might be benefit in creating a large-scale simulated dataset (maybe even with simulated baseline and/or MMs) to see how well the LCM algorithms can separate narrow from broad signals. I’ve seen some large simulated datasets used in recent deep-learning quantification papers…


I agree that synthetic data can help in evaluating the performance of various fitting tools. Short term, I do not plan on developing another version of fitting challenge. Three of us who put it together learned a lot from this experience so long term, we might use this experience.