Validation of IXIQ.Ai+gBSI : an automated framework for caudate atrophy measurement in Huntington’s disease
We have developed a fully automated framework that uses deep learning for caudate segmentation (IXIQ.Ai) and generalised BSI (gBSI) for longitudinal measurements.
Here, we validate the new method by comparing its volumetric scores with those of the standard manual pipeline (Man+BSI). Man+BSI produced larger caudate volumes than IXIQ.Ai+gBSI at baseline, but there was no significant difference in annualized volume change.
In addition, both methods showed significant differences in baseline volume and longitudinal change between symptomatic and non-symptomatic HD gene-carriers and controls, with comparable effect sizes. The IXIQ.Ai+gBSI method therefore produces volume estimates that are equally sensitive to disease progression as the Man+BSI, but in a faster, automated fashion.