MaPPeRTrac: A Massively Parallel, Portable, and Reproducible Tractography Pipeline

Area of Science:

• Neuroimaging
• Computational Neuroscience
• High-Performance Computing

Background:

• Diffusion MRI tractography is computationally intensive and requires complex software orchestration.
• Current methods present significant challenges for large-scale connectome analysis.

Purpose of the Study:

• To develop an automated, efficient, and user-friendly tractography pipeline.
• To simplify and accelerate the generation of structural connectomes from MRI data.
• To enhance the accessibility of large-scale brain connectome research.

Main Methods:

• Developed MaPPeRTrac, an edge-centric tractography pipeline for automated probabilistic or deterministic tractography.
• Containerized dependencies using Singularity (Apptainer) and organized data with Brain Imaging Data Structure (BIDS).
• Leveraged the Parsl parallel programming framework for efficient utilization of High-Performance Computing (HPC) resources.

Main Results:

• MaPPeRTrac fully automates the process from MRI data to edge density images (EDI) of structural connectomes.
• The pipeline demonstrates significant acceleration and simplification of tractography on diverse HPC environments.
• Enabled the creation of connectome datasets of unprecedented size.

Conclusions:

• MaPPeRTrac streamlines and accelerates diffusion MRI tractography, overcoming previous computational and dependency challenges.
• The pipeline's design promotes FAIR data principles and broadens access to advanced connectome research.
• Public availability and cross-platform compatibility facilitate wider adoption in the neuroimaging community.