Lab news and blog

29th May 2020 - Adaptive Baseline Fitting (ABfit)

Very pleased to announce the acceptance of my new MR Spectroscopy analysis method "Adaptive Baseline Fitting" (ABfit), currently in press at Magnetic Resonance in Medicine: link to preprint. A badly structured MRS analysis software ramble follows.

In my opinion, MRS fitting algorithms haven't improved in a significant way since the LCModel method was published, a staggering 27 years ago, and this has been both a blessing and a curse for MRS. A blessing because we've had access to a robust and highly optimised method for an long time, and it has rightly become the standard approach in the literature. A curse because the method is proprietary, relatively costly ($20k), and crucially, key algorithmic details are missing from the original paper - or terse to the point of obfustication.

Whilst a debate over pros and cons of proprietary vs open source software and data is a lengthy one, it's hard to argue that fMRI has suffered from having three excellent and open analysis packages: AFNI, FSL and SPM. This initial culture of openness (likely thanks to Karl Friston for choosing to release SPM under the GPL licence) has led to important initiatives such as the NIfTI file format (MRS file formats are an absolute mess), OpenNeuro, BIDS and many others.

Whilst I think ABfit is extremely useful as an MRS analysis method, its greater value is having an accompanying paper with a detailed explanation of the theory and justification for each stage of the algorithm. The method is also open source, and all code/data is freely available to fully reproduce the results of the paper github link. Finally, ABfit is integrated into a comprehensive MRS analysis package "SPectroscopy ANalysis Tools" spant, written in R, to aid the development of custom processing pipelines, batch analyses, data visualization and statistical interpretation.

The hope is that open and transparent methods encourages greater scrutiny and acknowledgment of their limitations - a necessary requirement for improvement. I have seen this in happening in the fMRI world, from the perspective of an interested outsider, and I think MRS would benefit greatly from a similar culture. I've played a small part to promote this type of activity in the past through the TARQUIN software and users' group, however I'm pleased to see bigger and better efforts currently underway thanks to Georg Oeltzschner, Brian Soher and Will Clarke's creation of the MRSHub. Let's hope that fostering a culture of openness and transparency in MRS research will give it the final push needed to become an essential clinical tool.

Martin Wilson