Using Lee et al. 2018 pediatric water relaxation values with Osprey RelaxationAtlas?

[aayrolles]

Hello all,
I am working with pediatric MRS data (children / early adolescents, e.g. ~8–15 years). I understand that the current RelaxationAtlas implementation is based on linear models derived from an adult cohort (20–60 years)

I would like to ask for your opinion on whether it would make sense (methodologically) to use pediatric water relaxation values from Lee et al., 2018 (NMR in Biomedicine, DOI: 10.1097/RLI.0000000000000435) within Osprey for younger subjects, instead of extrapolating the adult-based atlas to ages below the training range.

More specifically:
Do you see any major conceptual or technical issues with overriding the default water T1/T2 (and proton density) values in Osprey using pediatric literature values (e.g. Lee 2018) for children/adolescents?
Would you consider this preferable to extrapolating the adult RelaxationAtlas to pediatric ages?
Is there a other recommended way within Osprey to work with pediatric MRI?

I am not expecting Osprey to natively support Lee 2018, but I would appreciate your guidance on best practice from a methodological standpoint.

Thank you very much for your time and for making Osprey openly available :)

Anaël Ayrolles

[schorschinho]

Hi @aayrolles,

Thanks for your message! It would absolutely make sense to use pediatric/adolescent water relaxation values from this paper.

Indeed, I also wouldn't extrapolate the linear trends from our 20-60 relaxation atlas into younger age. One interesting thing you could do is look whether the predictions from the two papers approximately agree at an age of around 20.

I currently don't have the bandwidth to implement this, but you should be able to do it relatively easily by modifying OspreyQuantify - simply introduce a new option for MRS.opts.quantify (say, PediatricRelaxCorr) and branch out in lines 322f., introduce a new function (you could name it quantTissLee) that accepts subject age as an input and then applies the various models from the paper to calculate T1 and T2.

If you make it this far, I'd appreciate a pull request so we can merge it into OspreyQuantify (obviously with attribution).

Best,
Georg