Disentangling axonal loss and demyelination using multi‐modal imaging: Application to young onset Alzheimer’s disease

Background Mapping axonal loss in young‐onset Alzheimer’s disease (YOAD) can provide a clearer picture of neurodegenerative processes. Axonal loss has been estimated using metrics derived from diffusion MRI (dMRI). However, these measures are also sensitive to myelin, thus cannot unambiguously disentangle axon loss and demyelination. Here, we address this with multi‐modal imaging, combining dMRI with magnetisation transfer saturation (MTsat) MRI, a technique sensitive specifically to myelin. Method Neurite density index (NDI), a dMRI estimate from NODDI, relates to axon volume fraction (AVF), an axon density estimate unconfounded by myelin: AVF=(1‐MVF)(1‐FWF)NDI (Stikov et al. Neuroimage ; 118:397‐405, 2015 ). MVF is the myelin volume fraction, an estimate of myelin density, and FWF the free water fraction from NODDI. MVF is estimated from MTsat imaging (Mohammadi et al. Frontiers in Neuroscience ; 9, 9:441, 2015). Maps of AVF, MVF and NDI were computed for 21 healthy controls and 30 YOAD patients who underwent NODDI‐compatible dMRI and MTsat MRI. Tract‐based spatial statistics (TBSS) was used to map significant changes ( p <0.05, FWE‐corrected) along white matter (WM) skeleton (Fig. 2, lower) of the population template. Result Figure 1 shows the population‐averaged AVF and MVF of each point along the WM skeleton, demonstrating NDI is influenced by both AVF and MVF. While AVF and NDI correlate strongly (R 2 =0.46, p <5e‐10), suggesting NDI is primarily a marker of axon density, MVF also contributes to the observed NDI variation (R 2 =0.16, p<5e‐10). Figure 2 shows TBSS group differences in these metrics. MVF reductions in fronto‐occipital regions, indicating demyelination, overlap with NDI reductions but not AVF changes (Fig. 2A). AVF reductions in the splenium, indicating axonal loss, did not overlap with NDI changes (Fig. 2B). This demonstrates sub‐optimal specificity and sensitivity of NDI to axonal loss. Most AVF reductions; in the posterior tracts, fornix and the corpus callosum splenium, areas associated with symptoms of YOAD; were accompanied by decreased NDI. Concomitant MVF reductions occurred in posterior areas (Fig 2C). Conclusion This study presents a multi‐modal imaging approach to disambiguate markers of axonal loss and demyelination, effects previously entangled in single modality DWI analysis. This technique may provide new insight into in vivo pathological processes in neurodegenerative diseases.