Normalization of MRI T1w between-scan effects for improved longitudinal volumetric estimates

Abstract

Variations in image intensities and contrasts between magnetic resonance imaging (MRI) acquisitions affect the subsequent image processing and its derived outcomes. Therefore, comparability between acquisitions is improved if we reduce these variations. This is especially relevant for longitudinal studies where a change of scanner or acquisition protocol often happens between subsequent examinations. In this study, we use a robust intensity distribution alignment (RIDA) method to reduce between-scan effects and improve longitudinal volume change estimates between two MRI scanners - Siemens 1.5T Avanto and 3T Skyra. The method is based on MRI T1w images acquired in close succession and robustly aligns two cumulative distribution functions of voxel intensities to harmonize image intensities and improve image-derived outcomes of a range of subcortical brain. We compare RIDA with volume-based correction - a simple linear regression model. In both cases, we derive intensity and volume transformations from a training dataset of 20 participants scanned on both scanners on the same day and apply to an independent longitudinal test dataset of 243 participants. All participants in the test set were scanned at the Avanto scanner at the baseline and then at the Avanto and Skyra scanners on the same day at the follow-up, on average 4.4 years (sd = 0.5 years) later. This allowed us to directly assess the effect of scanner and protocol change on the longitudinal change estimates. Eight subcortical brain regions were segmented using SAMSEG, and annualized symmetrized percent change in volume between time points was calculated. We find that RIDA significantly reduces between-scan effects and improves longitudinal volume estimates for the amygdala and lateral ventricles. It also reduces between-scan effects for caudate, putamen, and thalamus, but not as much as linear regression models. Whether the method will be useful for a particular study will depend on the image intensity profiles of the scans. Therefore, a pilot study of double-scanned participants is recommended to assess the advantages of the method for the analysis in question.