Detecting Huntington's disease before symptoms appear has long been a clinical priority, given that neurodegeneration begins years prior to any observable motor or cognitive decline. A new imaging approach combining multiple MRI parameters with a motion-resistant acquisition sequence suggests it may be possible to classify disease stage — including pre-symptomatic carriers — with meaningful accuracy, potentially reshaping how clinicians monitor at-risk individuals and evaluate neuroprotective interventions.
The study enrolled 22 HD gene carriers, split between 14 pre-manifest (Pre-HD) and 8 manifest (M-HD) patients, alongside 27 healthy controls. Rather than relying on a single imaging contrast, investigators used the MOLED sequence — an ultrafast acquisition technique that suppresses motion artifacts — to simultaneously generate T2 and T2* relaxometry maps alongside conventional structural MRI at 3.0 Tesla. Eight deep brain regions, primarily within the basal ganglia, were analyzed using histogram-derived texture features (variance, skewness, maximum) rather than simple regional means. When machine learning classifiers combined volumetric structural data with these parametric mapping features, classification performance reached its highest point — notably outperforming any single-modality approach alone.
The broader significance lies in what this framework implies for neurodegenerative disease monitoring generally. Striatal atrophy in HD is well-documented, but mean-volume measures often lack the sensitivity to capture early or subtle microstructural disintegration. Histogram-based texture analysis extracts distributional information across voxels, capturing heterogeneity that averaged metrics obscure — an approach gaining traction in multiple sclerosis and Parkinson's research as well. The MOLED sequence's cross-scanner reproducibility is a practical strength, addressing a longstanding barrier to multicenter neuroimaging trials. Key limitations include a small cohort (n=49 total), a cross-sectional design that cannot confirm longitudinal sensitivity to disease progression, and the absence of independent external validation of the machine learning models. This is incremental but methodologically substantive work, laying groundwork for larger prospective biomarker validation studies.