Young Onset Alzheimer's disease (YOAD), defined as symptom onset less than 65 years, often present diverse syndromes. Its syndromic diversity may be underpinned by different patterns of network breakdown^1-5, suggesting that studying changes in white matter (WM) structural connections in this population can be especially pertinent. Previous diffusion tensor imaging (DTI) studies in patients with YOAD showed decreased fractional anisotropy (FA) and increased diffusivity in WM of patients^6-8. Whilst sensitive to microstructure changes in WM, DTI indices are not specific to different tissue properties (e.g., neurite density, which might be one of the main changes during the neurodegenerative process). Neurite orientation dispersion and density imaging (NODDI)⁹ is a novel diffusion MRI technique that provides more tissue specific information than DTI. The technique estimates the key microstructural factors contributing to the DTI indices, including neurite density, neurite orientation dispersion and fraction of free water. The aims of this study are to investigate the potential utility of the NODDI indices for detecting tissue specific WM changes in YOAD and for exploring whether unique patterns of WM changes exist in YOAD subtypes.

Subjects: 33 YOAD patients (age:60.6±5.1yrs, male/female:14/19, age at symptom-onset:55.0±4.1yrs) and 21 controls (age:60.7±5.7yrs, male/female:10/11) were scanned on the same 3T Siemens Trio scanner at Queen Square, University College London. The patient group consists of two syndromic variants: 8 individuals with Posterior Cortical Atrophy (PCA)¹⁰ and 25 with typical memory-led AD (tAD)¹¹.

Diffusion acquisition: We used a three-shell diffusion sequence optimised for NODDI (64, 32, and 8 diffusion-weighted directions at b=2000, 700 and 300 s/mm²; 14 b=0 images; 55 slices; voxel size 2.5x2.5x2.5 mm3; TR/TE=7000/92 ms; Acquisition Time 15 mins). DTI dataset was acquired using a single-shot, spin-echo echo planar imaging sequence (64 diffusion-weighted directions at b=1000 s/mm²; 9 b=0 images; 55 slices; voxel size 2.5x2.5x2.5 mm³; TR/TE = 6900/91ms; Acquisition Time 16.5 mins).

Image Processing: After correcting motion and eddy-current distortion, we fitted NODDI and tensor model using NODDI Matlab toolbox¹² and FSL¹³. We used DTI-TK¹⁴ to create a population-specific tensor template and spatially normalize all the parameter maps to the template.

Statistical Analysis: Tract-Based Spatial Statistics (TBSS)¹⁵ was used to detect group differences in NODDI and DTI indices between tAD and controls as well as between PCA and controls. To ensure a fair comparison between syndromic variants, one additional analysis was carried out between a subset of 8 tAD patients, that are age-and-gender matched to 8 PCA patients, and controls. For all the group comparisons, we performed 5000 permutations and included age and gender as covariates, and corrected multiple comparisons using Threshold-Free Cluster Enhancement (TFCE).

1) WM changes in tAD compared with controls

DTI findings: There was no significant change in FA at p<0.01, but at a less strict threshold of p<0.05 decreased FA were found in the tAD group posteriorly (Figure 1). Additionally, tAD patients had increased axial diffusivity (AxD) in central WM at p<0.01 and increased radial diffusivity (RD) in predominantly, but not exclusively, more posterior WM at p<0.05.

NODDI findings: The tAD group had widespread reduction in neurite density index (NDI) at p<0.01, in the corpus callosum, temporo-occipital, parietal and frontal WM tracts. There was also decreased orientation dispersion index (ODI) in large central WM tracts, e.g., corpus callosum at p<0.01, consistent with the AxD findings. Free water component increased in the corpus callosum at p<0.05; no significant changes survived at p<0.01. The complementary information provided by NODDI allows for a more fine-grained understanding of the conventional changes seen in DTI studies on neurodegenerative diseases. For example, NODDI reveals that the regions with reduced FA are underpinned by decreased NDI (Figure 2a). The technique can also identity WM changes in the regions where FA is not perturbed from a combination of reduced NDI and reduced ODI (Figure 2b).

2) NODDI metrics in YOAD syndromic variants

In the tAD subset, we found similar widespread decreased NDI than controls at p<0.01 (Figure 3a). Whereas, compared to controls, PCA group showed more focal decreased NDI in the posterior areas at p<0.01 (Figure 3b), and more focal ODI changes in posterior areas (Figure 3d) in contrast to the group differences between tADs and controls (Figure 3b). When comparing tADs and PCAs directly, no significant difference in NODDI parameters was found, which might be due to the small numbers in each group.

We showed that NODDI metrics not only enable tissue-specific microstructure characterization in YOAD but also provide enhanced sensitivity over DTI measures. We additionally demonstrated that NODDI can be used to reveal distinct profiles of WM damage in YOAD syndromic subgroups.