Comparison of diffusion metrics in dementia

Abstract

AbstractBackgroundThe recent FDA approval of a disease‐modifying drug for Alzheimer’s disease has stressed the need for a test which can be used at scale to detect neurodegeneration. MRI diffusion shows great promise (Douaud et al., 2013), and recently a multiplicity of analysis methods have been developed. We look at which diffusion analysis method might be most sensitive.MethodWe worked with data from Alzheimer’s Disease Neuroimaging Initiative (ADNI) (Weber et al., 2021). Participants with Alzheimer’s disease (AD) and a group of age‐matched healthy controls (HC) were identified. T1 images were processed with FreeSurfer v7 (Fischl, 2012), providing tissue segmentations and cortical parcellations, as well as regional volumes. Diffusion images were preprocessed using the PreQual pipeline in native space (including FSL’s Topup and Eddy) (Cai et al, 2021), and analysed with Diffusion Tensor Imaging (DTI), Neurite Orientation Dispersion and Density Imaging (NODDI) (Zhang, Schneider, et al, 2012), Apparent Measures Using Reduced Acquisitions (AMURA) (Aja‐Fernández et al, 2022), Diffusion Orientation Complexity (DOC) (Haroon et al, 2009) and Probabilistic Index of Connectivity (PICo) (Parker et al, 2003). Regional values were extracted in FreeSurfer regions (grey and white matter).Pairwise Mann‐Whitney U‐tests were carried out between AD and HC regional measures. Regions with the most statistically significant volume differences were selected; diffusion‐based metrics for these regions were calculated using weighted values and compared. The 10 metrics with the smallest p‐values were used to find the most significant regions, and these were compared with volumes.ResultParticipants: 100 AD (38 female, age 74.5±7.9 years) and 207 HC (132 female, age 72.0±7.4 years).We selected the 25 most significant regions from the volumetric analyses. From these regions, the most effective diffusion metrics were DTI‐derived mean, axial and radial diffusivities, and 5 of the AMURA metrics.Given the 10 most significant metrics, of the 15 most significant volumes, 8 were also in the top diffusion regions.ConclusionThis study demonstrates that DTI metrics are most sensitive at differentiating HC from AD. If confirmed this has clinical significance, as these scans are widely available.