Microstructure.jl: A Julia package for probabilistic microstructure model fitting with diffusion MRI

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Biophysics

Background:

• Diffusion MRI and diffusion-relaxometry are crucial for non-invasively probing tissue microstructure.
• Accurate estimation of microstructural parameters requires sophisticated modeling and robust fitting techniques.
• Existing tools may lack flexibility or comprehensive uncertainty quantification for advanced analyses.

Purpose of the Study:

• Introduce Microstructure.jl, a Julia package for probabilistic estimation of tissue microstructural parameters.
• Provide a flexible framework for defining compartment models and implementing advanced estimation methods.
• Demonstrate the package's utility in optimizing acquisition protocols and analyzing in vivo/ex vivo MRI data.

Main Methods:

• Developed Microstructure.jl in Julia, offering a flexible and extensible framework.
• Implemented established models like Spherical Mean Technique (SMT), Standard Model Imaging (SMI), and Soma and Neurite Density Imaging (SANDI).
• Integrated probabilistic estimation methods including Markov Chain Monte Carlo (MCMC) and neural network-based Monte Carlo dropout.

Main Results:

• Demonstrated Microstructure.jl's capability in optimizing acquisition protocols using synthesized data.
• Evaluated fitting performance for axon diameter index estimation.
• Successfully applied the package to analyze in vivo human connectome project data (SMT) and high-gradient scanner data (SANDI), and ex vivo Prisma scanner data (SMI with free water).

Conclusions:

• Microstructure.jl provides a powerful, unified, and flexible platform for microstructural parameter estimation and uncertainty quantification.
• The package supports various MRI acquisition types and scanner capabilities, applicable to both in vivo and ex vivo data.
• It facilitates advanced neuroimaging research by enabling robust analysis of complex tissue microstructures.