arXiv subjects

Chris A. Clark

Publications and source records attributed to Chris A. Clark.

3 recordsLinked to original sources

arXiv2024

Spherical convolutional neural networks can improve brain microstructure estimation from diffusion MRI data

Diffusion magnetic resonance imaging is sensitive to the microstructural properties of brain tissue. However, estimating clinically and scientifically relevant microstructural properties from the measured signals remains a highly challenging inverse problem that machine learning may help solve. This study investigated if recently developed rotationally invariant spherical convolutional neural networks can improve microstructural parameter estimation. We trained a spherical convolutional neural network to predict the ground-truth parameter values from efficiently simulated noisy data and applied the trained network to imaging data acquired in a clinical setting to generate microstructural parameter maps. Our network performed better than the spherical mean technique and multi-layer perceptron, achieving higher prediction accuracy than the spherical mean technique with less rotational variance than the multi-layer perceptron. Although we focused on a constrained two-compartment model of neuronal tissue, the network and training pipeline are generalizable and can be used to estimate the parameters of any Gaussian compartment model. To highlight this, we also trained the network to predict the parameters of a three-compartment model that enables the estimation of apparent neural soma density using tensor-valued diffusion encoding.

eess.IV↗

arXiv2022

Improved reproducibility of diffusion kurtosis imaging using regularized non-linear optimization informed by artificial neural networks

Diffusion kurtosis imaging is an extension of diffusion tensor imaging that provides scientifically and clinically valuable information about brain tissue microstructure but suffers from poor robustness to noise, especially in voxels containing tightly packed aligned axons. We present a new algorithm for estimating diffusion and kurtosis tensors using regularized non-linear optimization and make it publicly available in an easy-to-use open-source Python software package. Our approach uses fully-connected feed-forward neural networks to predict kurtosis values in voxels where the standard non-linear least squares fit fails. The predicted values are then used in the objective function to avoid implausible kurtosis values. We show that our algorithm is more robust than standard non-linear least squares and a previously proposed regularized non-linear optimization method. The algorithm was then applied on a multi-site scan-rescan dataset acquired using a clinical scan protocol to assess the reproducibility of diffusion kurtosis parameter estimation in human white matter using the proposed algorithm. Our results show that the reproducibility of diffusion kurtosis parameters is similar to diffusion tensor parameters.

physics.med-ph↗

arXiv2019

Validation and noise robustness assessment of microscopic anisotropy estimation with clinically feasible double diffusion encoding MRI

Purpose: Double diffusion encoding (DDE) MRI enables the estimation of microscopic diffusion anisotropy, yielding valuable information on tissue microstructure. A recent study proposed that the acquisition of rotationally invariant DDE metrics, typically obtained using a spherical "5-design", could be greatly simplified by assuming Gaussian diffusion, facilitating reduced acquisition times that are more compatible with clinical settings. Here, we aim to validate the new minimal acquisition scheme against the standard DDE 5-design, and to quantify the proposed method's noise robustness to facilitate future clinical use. Methods: DDE MRI experiments were performed on both ex vivo and in vivo rat brains at 9.4 T using the 5-design and the proposed minimal design and taking into account the difference in the number of acquisitions. The ensuing microscopic fractional anisotropy (μFA) maps were compared over a range of b-values up to 5000 s/mm2. Noise robustness was studied using analytical calculations and numerical simulations. Results: The minimal protocol quantified μFA at an accuracy comparable to the estimates obtained via the more theoretically robust DDE 5-design. μFA's sensitivity to noise was found to strongly depend on compartment anisotropy and tensor magnitude in a non-linear fashion. When μFA < 0.75 or when mean diffusivity is particularly low, very high signal to noise ratio (SNR) is required for precise quantification of μFA. Conclusion: Our work supports using DDE for quantifying microscopic diffusion anisotropy in clinical settings but raises hitherto overlooked precision issues when measuring μFA with DDE and typical clinical SNR.

physics.med-ph↗