purple-mri

purple-mri (Penn Utilities for Registration and Parcellation of Ex Vivo MRI) is a computational framework for segmentation, registration, surface reconstruction, and atlas-based parcellation of ultra–high-resolution postmortem human brain MRI.

The toolkit is designed for ex vivo whole-hemisphere MRI acquired at submillimeter resolution (often <300µm) at ultra-high field strengths (e.g., 7T), where conventional in vivo pipelines fail due to fixation-driven contrast shifts, specimen-specific geometry, and extreme spatial resolution.

purple-mri integrates deep learning–based voxel segmentation with classical surface-based modeling to produce topology-stable cortical reconstructions and native-space atlas parcellations suitable for vertex-wise and group-level morphometric analysis.

_images/animation.gif

Introductory Video

Core Capabilities

purple-mri enables:

  • Automated multi-label tissue segmentation of postmortem MRI

  • Topology-aware cortical ribbon refinement

  • Native-space surface reconstruction

  • Surface-based parcellation using established atlases (e.g., DKT, Schaefer, Glasser, von Economo–Koskinas)

  • Ex vivo ↔ in vivo volumetric registration using classical optimization methods and modern deep learning-based ones

  • Intensity-based population-specific volumetric template construction

  • Vertex-wise statistical modeling (e.g., GLM analyses in template space) allowing the integration of morphometric measures with external biological variables

_images/purple_flowchart_UPDATED.png

Scientific Context

We have applied our tools across a spectrum of ultra-high-resolution multi-modal ex vivo MRI spanning from developmental disorders (Sudden Infant Death Syndrome) in infants to neurodegenerative diseases (Alzheimer’s disease and related dementias) in adults. Our toolkit purple-mri has enabled systematic analysis of ultra-high-resolution postmortem MRI linking pathology to in vivo MRI via ex vivo MRI to allow discovery of region-specific morphometry–pathology signatures that can inform the development of disease-specific in vivo biomarkers.

Getting Started

Workflows

Reference

_images/python.png _images/pip.png _images/docker.png _images/singularity.png _images/fs.png _images/itksnap.png