Abstract:Abstract In diffusion MRI analysis, advances in biophysical multi-compartment modeling have gained popularity over the conventional Diffusion Tensor Imaging (DTI), because they possess greater specificity in relating the dMRI signal to underlying cellular microstructure. Biophysical multi-compartment models require parameter estimation, typically performed using either Maximum Likelihood Estimation (MLE) or using Monte Carlo Markov Chain (MCMC) sampling. Whereas MLE provides only a point estimate of the fitted model parameters, MCMC recovers the entire posterior distribution of the model parameters given the data, providing additional information such as parameter uncertainty and correlations. MCMC sampling is currently not routinely applied in dMRI microstructure modeling because it requires adjustments and tuning specific to each model, particularly in the choice of proposal distributions, burn-in length, thinning and the number of samples to store. In addition, sampling often takes at least an order of magnitude more time than non-linear optimization. Here we investigate the performance of MCMC algorithm variations over multiple popular diffusion microstructure models to see whether a single well performing variation could be applied efficiently and robustly to many models. Using an efficient GPU-based implementation, we show that run times can be removed as a prohibitive constraint for sampling of diffusion multi-compartment models. Using this implementation, we investigated the effectiveness of different adaptive MCMC algorithms, burn-in, initialization and thinning. Finally we apply the theory of Effective Sample Size to diffusion multi-compartment models as a way of determining a relatively general target for the number of samples needed to characterize parameter distributions for different models and datasets. We conclude that robust and fast sampling is achieved in most diffusion microstructure models with the Adaptive Metropolis-Within-Gibbs (AMWG) algorithm initialized with an MLE point estimate, in which case 100 to 200 samples are sufficient as a burn-in and thinning is mostly unnecessary. As a relatively general target for the number of samples, we recommend a multivariate Effective Sample Size of 2200.

Evaluation of the Influence of Data Sampling Schemes on Neural Diffusion Models

Analysis of the effects of noise, DWI sampling, and value of assumed parameters in diffusion MRI models

Angular Resolution Enhancement of Diffusion MRI Data Using Inter-Subject Information Transfer.

Improving Estimation of Fiber Orientations in Diffusion MRI Using Inter-Subject Information Sharing

Novel Single and Multiple Shell Gradient Sampling Schemes for Diffusion MRI Using Spherical Codes

Multi-Site Harmonization of Diffusion MRI Data Via Method of Moments

Robustness of Quantitative Diffusion Metrics from Four Models: A Prospective Study on the Influence of Scan‐Rescans, Voxel Size, Coils, and Observers

Noise Reduction in Diffusion MRI Using Non-Local Self-Similar Information in Joint X-Q Space.

The Performance of NODDI Estimation Using a Common 2-Shell Protocol

Single- and Multiple-Shell Uniform Sampling Schemes for Diffusion MRI Using Spherical Codes

A deep learning–based method for improving reliability of multicenter diffusion kurtosis imaging with varied acquisition protocols

DIMOND: DIffusion Model OptimizatioN with Deep Learning

Effects of Non-Local Diffusion on Structural MRI Preprocessing and Default Network Mapping: Statistical Comparisons with Isotropic/anisotropic Diffusion.

Robust and fast Monte Carlo Markov Chain sampling of diffusion MRI microstructure models

Sub-DM:Subspace Diffusion Model with Orthogonal Decomposition for MRI Reconstruction

Evaluation of Optimized B-Value Sampling Schemas for Diffusion Kurtosis Imaging with an Application to Stroke Patient Data.

Sensitivity of quantitative diffusion MRI tractography and microstructure to anisotropic spatial sampling

More with less: Simulation-based inference enables accurate diffusion-weighted MRI with minimal acquisition time

Reliable Deep Diffusion Tensor Estimation: Rethinking the Power of Data-Driven Optimization Routine

Estimating Tissue Microstructure with Undersampled Diffusion Data via Graph Convolutional Neural Networks

Application of different diffuse gradient direction in human brain to diffusion-weighted image