Background TRACK-HD is an international multi-site study of premanifest and early Huntington's disease (HD) which aims to investigate a range of biomarkers for use in future therapeutic trials. Detecting subtle brain changes in slowly progressive heterogeneous diseases such as HD is a challenge. Manual techniques, the current ‘gold standard’, are time consuming, reliant on the rater skill and impractical for large scale use. Linear registration of serial MRI has successfully been used to detect whole brain and caudate atrophy in HD using the boundary shift integral (BSI). However, it is less robust when applied to structures without brain–CSF boundaries. Non-linear (fluid) registration allows assessment of within subject change across the entire brain so may be applicable to a wider array of structures, thus providing more regionally specific information.

Aims To develop and apply a fully automated technique sensitive to 1 year global and regional volumetric changes in premanifest and early HD subjects using fluid registration of serial MRI.

Methods/techniques T1 weighted 3T MR images were acquired at baseline and 1 year from 303 subjects (97 controls; 110 premanifest; 96 early HD) in the TRACK-HD study. Regions of interest (ROI) were generated on the baseline scans using the automated FSL software. Fluid registration was used to model within subject voxel level change over the scanning interval and the 1 year volume change was quantified by summing voxel change within each ROI.

Results/outcome Preliminary fluid derived results (whole brain and caudate) show good group discrimination, consistent with BSI findings. We will present fluid derived regional and global atrophy rates, group differences and a comparison of our data with BSI results where appropriate.

Conclusions We describe a technique which has the potential to provide fully automated global and regional measures of atrophy using multi site data. This may have utility as an outcome measure for future therapeutic trials in HD.