p-Brain: A Modular Open-Source Framework for Automated Quantitative DCE-MRI of Cerebral Perfusion, Microvasculature, and Blood-Brain Barrier Permeability.
Authors
Affiliations (2)
Affiliations (2)
- Functional Imaging Unit, Department of Clinical Physiology and Nuclear Medicine, Copenhagen University Hospital, Rigshospitalet, Glostrup, Denmark.
- Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark.
Abstract
We present p-Brain, a modular, open-source framework for reproducible, automated quantitative DCE-MRI at scale. Rather than a fixed pipeline, p-Brain is built from interchangeable stages (ingestion, <math xmlns="http://www.w3.org/1998/Math/MathML"> <mrow> <msub><mrow><mi>T</mi></mrow> <mrow><mn>1</mn></mrow> </msub> <mo>/</mo> <msub><mrow><mi>M</mi></mrow> <mrow><mn>0</mn></mrow> </msub> </mrow> </math> fitting, vascular and tissue ROI extraction, signal-to-concentration conversion, kinetic modeling, and quality control), each selected and configured through a single file-based interface, so any stage can be swapped or extended without modifying the surrounding code. In its default configuration, p-Brain converts signal to gadolinium concentration, derives arterial and venous input functions using convolutional neural network (CNN) slice selection and ROI segmentation, and produces voxelwise, regional, and whole-brain maps. It implements Patlak graphical analysis for the blood-brain barrier influx constant ( <math xmlns="http://www.w3.org/1998/Math/MathML"> <mrow> <msub><mrow><mi>K</mi></mrow> <mrow><mi>i</mi></mrow> </msub> </mrow> </math> ) and blood volume ( <math xmlns="http://www.w3.org/1998/Math/MathML"> <mrow> <msub><mrow><mi>v</mi></mrow> <mrow><mi>b</mi></mrow> </msub> </mrow> </math> ), and model-free Tikhonov-regularised residue deconvolution for cerebral blood flow (CBF), cerebral blood volume (CBV), and mean transit time (MTT), with structured metadata and stage-level quality-control artifacts for auditability. We validate p-Brain against an established reference workflow in two ways: On identical inputs its estimators reproduce the reference algorithms to machine precision, and as a fully automated pipeline it agrees with the reference voxelwise ( <math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mi>r</mi> <mo>></mo> <mn>0</mn> <mo>.</mo> <mn>96</mn></mrow> </math> , ICC <math xmlns="http://www.w3.org/1998/Math/MathML"><mrow><mo>></mo> <mn>0</mn> <mo>.</mo> <mn>96</mn></mrow> </math> ) across all five maps (CBF, CBV, MTT, <math xmlns="http://www.w3.org/1998/Math/MathML"> <mrow> <msub><mrow><mi>K</mi></mrow> <mrow><mi>i</mi></mrow> </msub> </mrow> </math> , <math xmlns="http://www.w3.org/1998/Math/MathML"> <mrow> <msub><mrow><mi>v</mi></mrow> <mrow><mi>b</mi></mrow> </msub> </mrow> </math> ) in 12 healthy controls. p-Brain runs on Linux, macOS, and Windows as a Python package and command-line tool, and is open and extensible to additional segmentation tools, input-function providers, and kinetic models.