Pooling quantitative MRI data: A multi-protocol study of healthy subcortical aging.

Quantitative MRI (qMRI) measures relaxation rates, exchange rates, and proton densities that reflect biophysical properties of tissue and are ideally free from protocol- and scanner-dependence. In practice, qMRI has not yet achieved this level of independence from sequence and hardware choice, and quantitative estimates often differ across sites and acquisition schemes. At the same time, pooling data across different sources can be beneficial to statistical power of longitudinal, cross-sectional, or case-control studies. Here, we investigate how protocol and hardware differences can affect pooling data from different sources in large ultra high field (UHF) qMRI studies in the context of healthy aging. We combine the openly available aging UHF qMRI MP2RAGEME-based dataset with two different multiparameter mapping (MPM)-based sets of qMRI data. We evaluate how pooling affects age dependence of qMRI parameters and investigate protocol-related biases, with a particular focus on subcortical structures. We focus the analysis, first, on replication and expansion of the previously published qMRI dataset on normative aging, second, the examination of the protocol influence on the estimated qMRI values, and third, on detecting the protocol effect on the age dependence inferred from the data. We find that the age-related changes in the pooled dataset for R1 measure around 4-17% of the lifespan mean in different structures. Similarly, age-related R2* variation in different structures constitutes around 6-30% of the lifespan mean in the pooled dataset. Subcortical structure volume change is on the order of 5-26%. We further observe larger relative difference between protocols for R1 and volume, while R2* remains more consistent for most regions. We show how pooling the UHF qMRI data from different sites and collected with different quantitative protocols can be both detrimental and beneficial for the analysis outcomes.