Likelihood-free Posterior Estimation and Uncertainty Quantification for Diffusion MRI models

Imaging Neuroscience, Год журнала: 2024, Номер 2, С. 1 - 22

Опубликована: Янв. 19, 2024

Abstract Diffusion magnetic resonance imaging (dMRI) allows to estimate brain tissue microstructure as well the connectivity of white matter (known tractography). Accurate estimation model parameters (by solving inverse problem) is thus very important infer underlying biophysical properties and fiber orientations. Although there has been extensive research on this topic with a myriad dMRI models, most models use standard nonlinear optimization techniques only provide an without any information (quantification) about uncertainty in their estimation. Further, effect derived microstructural measures downstream (e.g., fractional anisotropy) often unknown rarely estimated. To address issue, we first design new deep-learning algorithm identify number crossing fibers each voxel. Then, at voxel, propose robust likelihood-free deep learning method not mean multi-fiber biexponential model), but also its full posterior distribution. The distribution then used measures. We perform several synthetic in-vivo quantitative experiments demonstrate robustness our approach for different noise levels out-of-distribution test samples. Besides, computationally fast requires order magnitude less time than fitting techniques. proposed demonstrates much lower error (compared existing methods) estimating metrics, including orientation, tensor eigenvalues. methodology quite general can be from other model.

Язык: Английский

---

Soma and Neurite Density MRI (SANDI) of the in-vivo mouse brain and comparison with the Allen Brain Atlas

NeuroImage, Год журнала: 2022, Номер 254, С. 119135 - 119135

Опубликована: Март 23, 2022

Diffusion MRI (dMRI) provides unique insights into the neural tissue milieu by probing interactions between diffusing molecules and microstructure. Most dMRI techniques focus on white matter (WM) tissues, nevertheless, interest in gray characterizations is growing. The Soma Neurite Density (SANDI) methodology harnesses a model incorporating water diffusion spherical objects (assumed to be associated with cell bodies) impermeable "sticks" represent neurites), which potentially enables characterization of cellular neurite densities. Recognising importance rodents animal models development, aging, plasticity, disease, we here employ SANDI for in-vivo preclinical imaging provide first validation comparing metrics density reflected Allen mouse brain atlas. was implemented 9.4T scanner equipped cryogenic coil, experiments were carried out N = 6 mice. Pixelwise, ROI-based, atlas comparisons performed, magnitude vs. real-valued analyses compared, shorter acquisitions reduced number b-value shells investigated. Our findings reveal good reproducibility parameters, including sphere stick fractions, as well size (CoV < 7%, 12% 3%, respectively). Additionally, find very rank correlation SANDI-driven fraction contrast that represents density. We conclude viable technique can greatly contribute research

Язык: Английский

---

Improving MR cell size imaging by inclusion of transcytolemmal water exchange

NMR in Biomedicine, Год журнала: 2022, Номер 35(12)

Опубликована: Июль 7, 2022

The goal of the current study is to include transcytolemmal water exchange in MR cell size imaging using IMPULSED model for more accurate characterization tissue cellular properties (e.g., apparent volume fraction intracellular space vin ) and quantification indicators exchange. We propose a heuristic that incorporates into multicompartment diffusion-based method (IMPULSED) was developed previously extract microstructural parameters mean d assuming no For tdiff ≤ 5 ms, can be ignored, signal same as model. ≥ 30 we incorporated modified Kärger includes both restricted diffusion between compartments. Using simulations published vitro data, evaluated accuracy precision model-derived determined how they are dependent on SNR parameters. joint provides values sizes ranging from 10 12 microns when fast pre-exchange lifetime τin 100 ms) than IMPULSED, reduces bias IMPULSED-derived estimates , especially relatively slow > 200 ms). Indicators derived proposed linearly correlated with ground truth detect changes membrane permeability induced by saponin treatment murine erythroleukemia cancer cells. Our results suggest this not only improves parameters, but also usually ignored models tissues.

Язык: Английский

---

Estimating axial diffusivity in the NODDI model

NeuroImage, Год журнала: 2022, Номер 262, С. 119535 - 119535

Опубликована: Авг. 3, 2022

To estimate microstructure-related parameters from diffusion MRI data, biophysical models make strong, simplifying assumptions about the underlying tissue. The extent to which many of these are valid remains an open research question. This study was inspired by disparity between estimated intra-axonal axial diffusivity literature and that typically assumed Neurite Orientation Dispersion Density Imaging (NODDI) model (d

Язык: Английский

---

Diffusion MRI with pulsed and free gradient waveforms: Effects of restricted diffusion and exchange

NMR in Biomedicine, Год журнала: 2022, Номер 36(1)

Опубликована: Сен. 8, 2022

Monitoring time dependence with diffusion MRI yields observables sensitive to compartment sizes (restricted diffusion) and membrane permeability (water exchange). However, restricted exchange have opposite effects on the diffusion-weighted signal, which can lead errors in parameter estimates. In this work, we propose a signal representation that incorporates of both up second order b-value is compatible gradient waveforms arbitrary shape. The features mappings from waveform two scalars separately control sensitivity restriction exchange. We demonstrate these span two-dimensional space be used choose selectively probe or exchange, eliminating correlation between phenomena. found specific but unconventional shapes provide an advantage over conventional pulsed oscillating acquisitions. also show parametrization into understand protocols other approaches For example, variation mixing filter-exchange imaging corresponds our exchange-weighting scalar at fixed value restriction-weighting scalar. proposed was evaluated using Monte Carlo simulations identical parallel cylinders hexagonal random packing as well gamma-distributed radii. Results showed approach interval 4-12 μm rates simulated range 0 20 s-1 , there extracellular geometry. presented theory constitutes simple intuitive description how influence guide protocol design capable separating effects.

Язык: Английский

---

Higher order diffusion imaging as a putative index of human sleep-related microstructural changes and glymphatic clearance

NeuroImage, Год журнала: 2023, Номер 274, С. 120124 - 120124

Опубликована: Апрель 20, 2023

The brain has a unique macroscopic waste clearance system, termed the glymphatic system which utilises perivascular tunnels surrounded by astroglia to promote cerebrospinal-interstitial fluid exchange. Rodent studies have demonstrated marked increase in during sleep been linked sleep-induced expansion of extracellular space and concomitant reduction intracellular volume. However, despite being implicated pathophysiology multiple human neurodegenerative disorders, non-invasive techniques for imaging humans are currently limited. Here we acquired multi-shell diffusion weighted MRI (dwMRI) twenty-one healthy young participants (6 female, 22.3 ± 3.2 years) each scanned twice, once wakefulness induced combination one night deprivation 10 mg hypnotic zolpidem 30 min before scanning. To capture hypothesised sleep-associated changes intra/extracellular space, dwMRI were analysed using higher order modelling with prediction that increases interstitial (extracellular) volume would result decrease kurtosis, particularly areas associated slow wave generation at onset sleep. In line our hypothesis, observed global kurtosis (t15=2.82, p = 0.006) as well regional reductions early default mode network highly metabolically active wakefulness. Analysis higher-order representation (MAP-MRI) further indicated within domain rather than membrane permeability likely underpin kurtosis. These findings identify potential new method extend rodent suggest is also an humans.

Язык: Английский

---

Cellular Exchange Imaging (CEXI): Evaluation of a diffusion model including water exchange in cells using numerical phantoms of permeable spheres

Magnetic Resonance in Medicine, Год журнала: 2023, Номер 90(4), С. 1625 - 1640

Опубликована: Июнь 6, 2023

Purpose Biophysical models of diffusion MRI have been developed to characterize microstructure in various tissues, but existing are not suitable for tissue composed permeable spherical cells. In this study we introduce Cellular Exchange Imaging (CEXI), a model tailored cells, and compares its performance related Ball & Sphere (BS) that neglects permeability. Methods We generated DW‐MRI signals using Monte‐Carlo simulations with PGSE sequence numerical substrates made cells their extracellular space range membrane From these signals, the properties were inferred both BS CEXI models. Results outperformed impermeable by providing more stable estimates cell size intracellular volume fraction time‐independent. Notably, accurately estimated exchange time low moderate permeability levels previously reported other studies (). However, highly (), parameters less stable, particularly coefficients. Conclusion This highlights importance modeling quantify cellular substrates. Future should evaluate clinical applications such as lymph nodes, investigate potential biomarker tumor severity, develop appropriate account anisotropic membranes.

Язык: Английский

---

Low-field, high-gradient NMR shows diffusion contrast consistent with localization or motional averaging of water near surfaces

Magnetic Resonance Letters, Год журнала: 2023, Номер 3(2), С. 90 - 107

Опубликована: Апрель 9, 2023

Nuclear magnetic resonance (NMR) measurements of water diffusion have been extensively used to probe microstructure in porous materials, such as biological tissue, however primarily using pulsed gradient spin echo (PGSE) methods. Low-field single-sided NMR systems built-in static gradients (SG) much stronger than typical PGSE maximum strengths, which allows for the signal attenuation at extremely high b-values be explored. Here, we perform SG (SGSE) and stimulated (SGSTE) on cells, tissues, gels. Measurements fixed live neonatal mouse spinal cord, lobster ventral nerve starved yeast cells all show multiexponential a scale b with significant fractions observed × D0 ≫ 1 400 ms/μm2. These persistent trend surface-to-volume ratios these systems, expected from media theory. An exception found case vs. cords was attributed faster exchange or permeability millisecond timescale. Data suggests existence multiple processes neural may relevant modeling time-dependent gray matter. The multi-exponential is protons not macromolecules because it remains proportional normalized when specimen washed D2O. that persists also drastically reduced after delipidation, indicating originates lipid membranes restrict diffusion. stretched exponential character appears mono-exponential viewed (b×D0)1/3, suggesting originate localization motional averaging near sub-micron length scales. To try disambiguate two contributions, curves were compared varying temperatures. While align normalizing them scale, they separate scale. This supports source non-Gaussian displacements, but this interpretation still provisional due possible confounds heterogeneity, exchange, relaxation. types gel phantoms designed mimic extracellular matrix, one charged functional groups synthesized polyacrylic acid (PAC) another uncharged polyacrylamide (PAM), both exhibit 1, potentially interacting macromolecules. preliminary finding motivate future research into contrast mechanisms tissue low-field, high-gradient NMR.

Язык: Английский

---

Diffusion MRI with free gradient waveforms on a high-performance gradient system: Probing restriction and exchange in the human brain

NeuroImage, Год журнала: 2023, Номер 283, С. 120409 - 120409

Опубликована: Окт. 13, 2023

The dependence of the diffusion MRI signal on time carries signatures restricted and exchange. Here we seek to highlight these in human brain by performing experiments using free gradient waveforms that are selectively sensitive two effects. We examine six healthy volunteers both strong ultra-strong gradients (80, 200 300 mT/m). In an experiment featuring a large set with different sensitivities exchange (150 samples), our results reveal unique time-dependence grey white matter, where former is characterised latter predominantly exhibits diffusion. Furthermore, show independently varying can be used map brain. consistently find matter at least twice as fast across all subjects strengths. shortest times observed this study were cerebellar cortex (115 ms). also assess feasibility future clinical applications method work, grey-white contrast obtained 25-minute mT/m protocol preserved 4-minute 10-minute 80 protocol. Our work underlines utility for detecting due vivo, which may potentially serve tool studying diseased tissue.

Язык: Английский

---

The Larmor frequency shift of a white matter magnetic microstructure model with multiple sources

NMR in Biomedicine, Год журнала: 2024, Номер 37(8)

Опубликована: Март 29, 2024

Abstract Magnetic susceptibility imaging may provide valuable information about chemical composition and microstructural organization of tissue. However, its estimation from the MRI signal phase is particularly difficult as it sensitive to magnetic tissue properties ranging molecular macroscopic scale. The Larmor frequency shift measured in white matter (WM) depends on myelinated axons other magnetizable sources such iron‐filled ferritin. We have previously derived arising a dense medium cylinders with scalar arbitrary orientation dispersion. Here, we extend our model include microscopic WM anisotropy well spherical inclusions represent subcellular structures, biologically stored iron, so forth. validate analytical results computer simulations investigate feasibility estimating using simple iterative linear least squares without regularization or preconditioning. This done digital brain phantom synthesized diffusion measurements an ex vivo mouse at ultra‐high field.

Язык: Английский

---