Advanced Analysis of Diffusion Tensor Imaging Along With Machine Learning Provides New Sensitive Measures of Tissue Pathology and Intra-Lesion Activity in Multiple Sclerosis
Open Access
- 7 May 2021
- journal article
- research article
- Published by Frontiers Media SA in Frontiers in Neuroscience
Abstract
Tissue pathology in multiple sclerosis (MS) is highly complex, requiring multi-dimensional analysis. In this study, our goal was to test the feasibility of obtaining high angular resolution diffusion imaging (HARDI) metrics through single-shell modeling of diffusion tensor imaging (DTI) data, and investigate how advanced measures from single-shell HARDI and DTI tractography perform relative to classical DTI metrics in assessing MS pathology. We examined 52 relapsing-remitting MS patients who had 3T anatomical brain MRI and DTI. Single-shell HARDI modeling yielded 5 sub-voxel-based metrics, totalling 11 diffusion measures including 4 DTI and 2 tractography metrics. Based on machine learning of 3-dimensional regions of interest, we evaluated the importance of the measures through several tissue classification tasks. These included two within-subject comparisons: lesion versus normal appearing white matter (NAWM); and lesion core versus shell. Further, by stratifying patients as having high (above 75%ile) and low (below 25%ile) number of MS lesions, we also performed 2 classifications between subjects for lesions and NAWM respectively. Results showed that in lesion-NAWM analysis, HARDI orientation distribution function (ODF) energy, DTI fractional anisotropy (FA), and HARDI orientation dispersion index were the top three metrics, which together achieved 65.2% accuracy and 0.71 area under the receiver operating characteristic curve (AUROC). In core-shell analysis, DTI mean diffusivity (MD), radial diffusivity, and FA were the top three metrics, and MD dominated the classification, which achieved 59.3% accuracy and 0.59 AUROC alone. Between patients, FA was the leading feature in lesion comparisons, while ODF energy was the best in NAWM separation. Collectively, single-shell modeling of common diffusion data can provide robust orientation measures of lesion and NAWM pathology, and DTI metrics are most sensitive to intra-lesion abnormality. Combined analysis of both advanced and classical diffusion measures may be critical for improved understanding of MS pathology.Funding Information
- Alberta Innovates
- Multiple Sclerosis Society of Canada
This publication has 36 references indexed in Scilit:
- An automated tool for detection of FLAIR-hyperintense white-matter lesions in Multiple SclerosisNeuroImage, 2012
- Orientationally invariant indices of axon diameter and density from diffusion MRINeuroImage, 2010
- Diffusion imaging in multiple sclerosis: research and clinical implicationsNMR in Biomedicine, 2010
- Symmetric diffeomorphic image registration with cross-correlation: Evaluating automated labeling of elderly and neurodegenerative brainMedical Image Analysis, 2008
- Disability and T2 MRI lesions: a 20-year follow-up of patients with relapse onset of multiple sclerosisBrain, 2008
- Regularized, fast, and robust analytical Q‐ball imagingMagnetic Resonance in Medicine, 2007
- Diffusion tensor imaging of post mortem multiple sclerosis brainNeuroImage, 2007
- The Pathology of MSThe Neurologist, 2007
- Improved Optimization for the Robust and Accurate Linear Registration and Motion Correction of Brain ImagesNeuroImage, 2002