Recovering the T2 distribution from multi-echo T2 magnetic resonance (MR) signals is challenging but has high potential as it provides biomarkers characterizing the tissue micro-structure, such as the myelin water fraction (MWF). In this work, we propose to combine machine learning and aspects of parametric (fitting from the MRI signal using biophysical models) and non-parametric (model-free fitting of the T2 distribution from the signal) approaches to T2 relaxometry in brain tissue by using a multi-layer perceptron (MLP) for the distribution reconstruction. For training our network, we construct an extensive synthetic dataset derived from biophysical models in order to constrain the outputs with a priori knowledge of in vivo distributions. The proposed approach, called Model-Informed Machine Learning (MIML), takes as input the MR signal and directly outputs the associated T2 distribution. We evaluate MIML in comparison to a Gaussian Mixture Fitting (parametric) and Regularized Non-Negative Least Squares algorithms (non-parametric) on synthetic data, an ex vivo scan, and high-resolution scans of healthy subjects and a subject with Multiple Sclerosis. In synthetic data, MIML provides more accurate and noise-robust distributions. In real data, MWF maps derived from MIML exhibit the greatest conformity to anatomical scans, have the highest correlation to a histological map of myelin volume, and the best unambiguous lesion visualization and localization, with superior contrast between lesions and normal appearing tissue. In whole-brain analysis, MIML is 22 to 4980 times faster than the non-parametric and parametric methods, respectively.

Diffusion magnetic resonance imaging is a noninvasive imaging modality that has been extensively used in the literature to study the neuronal architecture of the brain in a wide range of neurological conditions using tractography. However, recent studies highlighted that the anatomical accuracy of the reconstructions is inherently limited and challenged its appropriateness. Several solutions have been proposed to tackle this issue, but none of them proved effective to overcome this fundamental limitation. In this work, we present a novel processing framework to inject into the reconstruction problem basic prior knowledge about brain anatomy and its organization and evaluate its effectiveness using both simulated and real human brain data. Our results indicate that our proposed method dramatically increases the accuracy of the estimated brain networks and, thus, represents a major step forward for the study of connectivity.

A new method substantially improves the accuracy of mapped brain networks using anatomy and microstructure informed tractography. Diffusion magnetic resonance imaging is a noninvasive imaging modality that has been extensively used in the literature to study the neuronal architecture of the brain in a wide range of neurological conditions using tractography. However, recent studies highlighted that the anatomical accuracy of the reconstructions is inherently limited and challenged its appropriateness. Several solutions have been proposed to tackle this issue, but none of them proved effective to overcome this fundamental limitation. In this work, we present a novel processing framework to inject into the reconstruction problem basic prior knowledge about brain anatomy and its organization and evaluate its effectiveness using both simulated and real human brain data. Our results indicate that our proposed method dramatically increases the accuracy of the estimated brain networks and, thus, represents a major step forward for the study of connectivity.

Graphical abstract

Non‐invasive axon diameter distribution (ADD) mapping using diffusion MRI is an ill‐posed problem. Current ADD mapping methods require knowledge of axon orientation before performing the acquisition. Instead, ActiveAx uses a 3D sampling scheme to estimate the orientation from the signal, providing orientationally invariant estimates. The mean diameter is estimated instead of the distribution for the solution to be tractable. Here, we propose an extension (ActiveAxADD) that provides non‐parametric and orientationally invariant estimates of the whole distribution.

Investigative studies of white matter (WM) brain structures using diffusion MRI (dMRI) tractography frequently require manual WM bundle segmentation, often called “virtual dissection.” Human errors and personal decisions make these manual segmentations hard to reproduce, which have not yet been quantified by the dMRI community. It is our opinion that if the field of dMRI tractography wants to be taken seriously as a widespread clinical tool, it is imperative to harmonize WM bundle segmentations and develop protocols aimed to be used in clinical settings. The EADC‐ADNI Harmonized Hippocampal Protocol achieved such standardization through a series of steps that must be reproduced for every WM bundle. This article is an observation of the problematic. A specific bundle segmentation protocol was used in order to provide a real‐life example, but the contribution of this article is to discuss the need for reproducibility and standardized protocol, as for any measurement tool. This study required the participation of 11 experts and 13 nonexperts in neuroanatomy and “virtual dissection” across various laboratories and hospitals. Intra‐rater agreement (Dice score) was approximately 0.77, while inter‐rater was approximately 0.65. The protocol provided to participants was not necessarily optimal, but its design mimics, in essence, what will be required in future protocols. Reporting tractometry results such as average fractional anisotropy, volume or streamline count of a particular bundle without a sufficient reproducibility score could make the analysis and interpretations more difficult. Coordinated efforts by the diffusion MRI tractography community are needed to quantify and account for reproducibility of WM bundle extraction protocols in this era of open and collaborative science.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Tractography is a unique tool to study neurological disorders for its ability to infer the major neural tracts from diffusion MRI data, thus allowing the investigation of the connectivity of the brain in-vivo. Tractography has seen a remarkable interest over the years and a large number of algorithms have been proposed. The choice of which method to use in a given application is usually a trade-off between its computational complexity and the quality of the reconstructions. So-called “shortest path” methods represent an interesting option, as they are computationally efficient, robust to noise and their formulation is very flexible. However, they also come with limitations. For instance, the reconstructed streamlines tend to be collapsed and to share part of their path, especially in regions with highly curved fiber bundles. This can introduce voxels with incorrectly high or low streamline density, which does not correspond to the underlying fiber geometry. To mitigate this problem, we propose an iterative procedure that uses microstructure information and provides feedback to the shortest path tractography algorithm about the plausibility of the reconstructions. We evaluated our method on a synthetic phantom and show that the spatial distribution of streamlines is in closer agreement with the ground truth.

Fiber tracking with diffusion‐weighted MRI has become an essential tool for estimating in vivo brain white matter architecture. Fiber tracking results are sensitive to the choice of processing method and tracking criteria.

Investigative studies of white matter (WM) brain structures using diffusion MRI (dMRI) tractography frequently require manual WM bundle segmentation, often called “virtual dissection”. Human errors and personal decisions make these manual segmentations hard to reproduce, which have not yet been quantified by the dMRI community. The contribution of this study is to provide the first large-scale, international, multi-center variability assessment of the “virtual dissection” of the pyramidal tract (PyT). Eleven (11) experts and thirteen (13) non-experts in neuroanatomy and “virtual dissection” were asked to perform 30 PyT segmentation and their results were compared using various voxel-wise and streamline-wise measures. Overall the voxel representation is always more reproducible than streamlines (≈70% and ≈35% overlap respectively) and distances between segmentations are also lower for voxel-wise than streamline-wise measures (¾3mm and ¾ûmm respectively). This needs to be seriously considered before using tract-based measures (e.g. bundle volume versus streamline count) for an analysis. We show and argue that future bundle segmentation protocols need to be designed to be more robust to human subjectivity. Coordinated efforts by the diffusion MRI tractography community are needed to quantify and account for reproducibility of WM bundle extraction techniques in this era of open and collaborative science.

Tractography is a family of algorithms that use diffusion-weighted magnetic resonance imaging data to reconstruct the white matter pathways of the brain. Although it has been proven to be particularly effective for studying non-invasively the neuronal architecture of the brain, recent studies have highlighted that the large incidence of false positive connections retrieved by these techniques can significantly bias any connectivity analysis. Some solutions have been proposed to overcome this issue and the ones relying on convex optimization framework showed a significant improvement. Here we propose an evolution of the Convex Optimization Modeling for Microstructure Informed Tractography (COMMIT) framework, that combines basic prior knowledge about brain anatomy with group-sparsity regularization into the optimization problem. We show that the new formulation dramatically reduces the incidence of false positives in synthetic DW-MRI data.

&NA; Spherical deconvolution methods are widely used to estimate the brain's white‐matter fiber orientations from diffusion MRI data. In this study, eight spherical deconvolution algorithms were implemented and evaluated. These included two model selection techniques based on the extended Bayesian information criterion (i.e., best subset selection and the least absolute shrinkage and selection operator), iteratively reweighted l2‐ and l1‐norm approaches to approximate the l0‐norm, sparse Bayesian learning, Cauchy deconvolution, and two accelerated Richardson‐Lucy algorithms. Results from our exhaustive evaluation show that there is no single optimal method for all different fiber configurations, suggesting that further studies should be conducted to find the optimal way of combining solutions from different methods. We found l0‐norm regularization algorithms to resolve more accurately fiber crossings with small inter‐fiber angles. However, in voxels with very dominant fibers, algorithms promoting more sparsity are less accurate in detecting smaller fibers. In most cases, the best algorithm to reconstruct fiber crossings with two fibers did not perform optimally in voxels with one or three fibers. Therefore, simplified validation systems as employed in a number of previous studies, where only two fibers with similar volume fractions were tested, should be avoided as they provide incomplete information. Future studies proposing new reconstruction methods based on high angular resolution diffusion imaging data should validate their results by considering, at least, voxels with one, two, and three fibers, as well as voxels with dominant fibers and different diffusion anisotropies. HighlightsThere is no single optimal SD method for all the different fiber configurations.Sparse algorithms to resolve fiber crossings with small inter‐fiber angles were found.Algorithms promoting more sparsity are less accurate in detecting smaller fibers.Future studies should validate their results by considering many fiber configurations.