Tractography and machine learning: Current state and open challenges

Poulin, Philippe; Jörgens, Daniel; Jodoin, Pierre-Marc; Descoteaux, Maxime

doi:10.1016/j.mri.2019.04.013

Cited by 60 publications

(55 citation statements)

References 70 publications

(160 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…C. Alexander et al 2017;Ghosh, Ianus, and Alexander 2018;D. S. Novikov et al 2019;Poulin et al 2019;Ravi et al 2019) The standard acquisition strategy for dMRI data is single diffusion encoding (SDE), which employs a pair of diffusion weighting gradients with identical areas, usually embedded before and after the refocusing pulse in a spin echo preparation, a sequence widely known also as pulsed gradient spin-echo (Stejskal and Tanner 1965). The SDE sequences are characterized by the gradient strength (G), duration (δ), time interval between the onset of the two gradients (Δ) and gradient orientation ( ").…”

Section: Introductionmentioning

confidence: 99%

On the generalizability of diffusion MRI signal representations across acquisition parameters, sequences and tissue types: chronicles of the MEMENTO challenge

Luca

Ianuş

Leemans

et al. 2021

Preprint

View full text Add to dashboard Cite

Diffusion MRI (dMRI) has become an invaluable tool to assess the microstructural organization of brain tissue. Depending on the specific acquisition settings, the dMRI signal encodes specific properties of the underlying diffusion process. In the last two decades, several signal representations have been proposed to fit the dMRI signal and decode such properties. Most methods, however, are tested and developed on a limited amount of data, and their applicability to other acquisition schemes remains unknown. With this work, we aimed to shed light on the generalizability of existing dMRI signal representations to different diffusion encoding parameters and brain tissue types. To this end, we organized a community challenge - named MEMENTO, making available the same datasets for fair comparisons across algorithms and techniques. We considered two state-of-the-art diffusion datasets, including single-diffusion-encoding (SDE) spin-echo data from a human brain with over 3820 unique diffusion weightings (the MASSIVE dataset), and double (oscillating) diffusion encoding data (DDE/DODE) of a mouse brain including over 2520 unique data points. A subset of the data sampled in 5 different voxels was openly distributed, and the challenge participants were asked to predict the remaining part of the data. After one year, eight participant teams submitted a total of 80 signal fits. For each submission, we evaluated the mean squared error, the variance of the prediction error and the Bayesian information criteria. Most predictions predicted either multi-shell SDE data (37%) or DODE data (22%), followed by cartesian SDE data (19%) and DDE (18%). Most submissions predicted the signals measured with SDE remarkably well, with the exception of low and very strong diffusion weightings. The prediction of DDE and DODE data seemed more challenging, likely because none of the submissions explicitly accounted for diffusion time and frequency. Next to the choice of the model, decisions on fit procedure and hyperparameters play a major role in the prediction performance, highlighting the importance of optimizing and reporting such choices. This work is a community effort to highlight strength and limitations of the field at representing dMRI acquired with trending encoding schemes, gaining insights into how different models generalize to different tissue types and fiber configurations over a large range of diffusion encodings.

show abstract

Section: Introductionmentioning

confidence: 99%

On the generalizability of diffusion MRI signal representations across acquisition parameters, sequences and tissue types: chronicles of the MEMENTO challenge

Luca

Ianuş

Leemans

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Future work will address the incorporation of structural priors from e.g. diffusion MRI [48] as well as the reformulation through recent, more sophisticated NN architectures (e.g. combinations of combination of ESN and LSTM models [49]) which have been built to facilitate training.…”

Section: Discussionmentioning

confidence: 99%

Echo State Network models for nonlinear Granger causality

Duggento

Guerrisi

Toschi

2019

Preprint

View full text Add to dashboard Cite

While Granger Causality (GC) has been employed in a large number of problems in network neuroscience, the vast majority of GC applications are based on linear multivariate autoregressive (MVAR) models. However, it is well known that real-life systems (and biological networks in particular) exhibit notable nonlinear behavior, hence undermining that validity of MVAR-based GC (MVAR-GC) approaches. Current nonlinear approaches to GC estimation only cater for additive nonlinearities or, alternatively, are based on recurrent neural networks (RNN) or Long short-term memory (LSTM) networks, which present considerable training difficulties and/or the need to be carefully tailored to specific applications. We define a novel approach to estimating nonlinear, directed within-network interactions based on a specific class of RNNs termed echo-state networks (ESN), where training is replaced by random initialization and the internal basis is built through orthonormal matrices. We reformulate the classical GC framework in terms of ESN-based models for arbitrarily complex networks, and characterize the ability of our ESN-based Granger Causality (ES-GC) estimator to capture nonlinear causal relations in a network of noisy Duffing oscillators, showing a net advantage of ES-GC in detecting nonlinear, causal links. We then explore the structure of ES-GC networks in the human brain employing functional MRI data from 1003 healthy subjects scanned at rest at 3T within the human connectome project, demonstrating the existence of previously unknown directed within-brain interactions. In summary, ES-GC performs better than commonly used and recently developed GC detection approaches, making it a valuable tool for the analysis of e.g. multivariate biological networks.

show abstract

“…It may be also attributed to the nonunique interpretations of the streamline topology within single voxel from the measured diffusion MRI signals at the voxel, causing a significant amount of false-positive connectivities 30 . In the past decades, tremendous efforts have been made to improving the quality of diffusion MR imaging 19,30,[32][33][34][35][36] and tractography algorithm [37][38][39] . The bias in the tractography algorithms and the ambiguity in the resulting streamline topology may be considerably further reduced or even resolved to a degree of satisfaction with the advances in diffusion MR imaging and computing technologies.…”

Section: Discussionmentioning

confidence: 99%

A Measure of Whole-Brain White Matter Topography

Wang¹,

Shi²

2019

Preprint

View full text Add to dashboard Cite

The unprecedentedly high-quality large-scale brain imaging datasets, from such as the Human Connectome Project (HCP) and UK-Biobank, provide a unique opportunity for measuring the white matter topography of the human brain. By leveraging the multi-shell diffusion MRI data from the original young adult HCP, we systematically develop a reliable measure of the whole-brain white matter topography, and we coin it topographic vector. As the main result, we find that the three most dominant dimensions of the topographic vectors strongly and linearly correlate with the coordinates of the corresponding streamlines of the whole-brain tractograms. Our results support the earlier prescient hypothesis that brain development follows a "base-plan" established by three (main) chemotactic gradients of early embryogenesis, and they implicate that the whole brain white matter tracts can be represented by vectors of a natural coordinate system. The topographic organisation, or topography, of the brain white matter was first measured using Klingler's brain dissection technique in early post-mortem studies 30 years ago 1 . However, our understanding of it is still limited. We view the topography of brain white matter pathways at two scales: local and global. The local topographic organisation, which is also known as the topographic regularity, has been defined as the preservation of the spatial relationship between neurons in point-to-point or region-toregion axonal connections 2,3 and it has been supported by tracer injection results 4-7 and diffusion magnetic resonance imaging (dMRI) studies 8-11 for various white matter

show abstract

Tractography and machine learning: Current state and open challenges

Cited by 60 publications

References 70 publications

On the generalizability of diffusion MRI signal representations across acquisition parameters, sequences and tissue types: chronicles of the MEMENTO challenge

On the generalizability of diffusion MRI signal representations across acquisition parameters, sequences and tissue types: chronicles of the MEMENTO challenge

Echo State Network models for nonlinear Granger causality

A Measure of Whole-Brain White Matter Topography

Contact Info

Product

Resources

About