DCEASY — Dynamic Contrast-Enhanced Analysis SYstem

DCEASY is a collection of open-source software maintained by the PET/MRI Lab for end-to-end Dynamic Contrast-Enhanced (DCE) MRI workflows — from DICOM-to-BIDS conversion and parametric mapping to pharmacokinetic modeling and arterial input function (AIF) detection. Choose the tool that fits your step in the pipeline below.

DCE Processing Overview

Run the whole pipeline at once

dce2bids AI-assisted DICOM → BIDS conversion — the required first step before either path below.

DCEPrep Automated & Dockerized — preprocessing through quality control and analysis (steps 1–5)

…or run it step by step

1 Convert and Preprocessing → standard MRI tools (dcm2niix, FSL, ANTs, etc.)
2 Generate T1 maps → parametric_scripts
3 Extract an AIF → AutoAIF (automatic) / AIFArtist (manual)
4 Signal intensity to concentration, fit pharmacokinetic models (Ktrans, ve, vp …) → ROCKETSHIP + Gpufit
5 Analyze & compare parametric maps → ROCKETSHIP

Tools

dce2bids

Python / AI

Start here. Converts raw DCE-MRI DICOMs straight from the scanner into a tidy, standard BIDS dataset. An AI coding agent (e.g. Claude Code) works out the right per-scanner settings once; after that each new participant converts with a single command. Lets the AI handle the required customization and trouble shooting, but includes guardrails and verification.

Auto-selects the dynamic DCE, VFA flip-angle, and structural series
AI assistant handles fiddly per-scanner/protocol setup, just once
Saves a reusable run_dce2bids.sh for future cases
DICOM → NIfTI with BIDS-compliant output and a status report
Built-in verify_bids.py validation

GitHub

ROCKETSHIP

MATLAB

A flexible, GUI-driven suite for full DCE-MRI analysis. Covers pre-contrast T1 mapping, AIF selection and fitting, multi-model pharmacokinetic curve fitting (Tofts, Extended Tofts, Patlak, Two-Compartment Exchange, FXR, …), and results visualization. Supports NIFTI and DICOM inputs and optional GPU acceleration.

Tofts, Extended Tofts, Patlak, 2CXM, FXR, tissue-uptake models
GUI-based workflow — no scripting required
Batch processing via parallel computing toolbox
Optional GPU acceleration (Gpufit)
Cited in BMC Medical Imaging

GitHub Wiki / Docs Paper (BMC 2015)

AutoAIF

Python

Preferred for AIF detection. A 3D U-Net deep-learning model (Keras/TensorFlow) that automatically detects the arterial input function in brain DCE-MRI. Pretrained weights are provided; the model handles multi-site data by resampling inputs to a canonical resolution and outputs a vascular function curve together with a 3D vascular mask.

Fully automatic — no manual ROI drawing required
Pretrained on multi-site brain DCE-MRI cohorts
Outputs vascular function curve + 3D mask (NIfTI)
Supports inference and fine-tuning on new datasets
Published in Magnetic Resonance in Medicine (2025)

GitHub Paper (MRM 2025)

AIFArtist

Python

A napari desktop application for manual arterial input function (AIF) annotation on 4D MRI NIfTI data. Designed for high-volume multi-rater review sessions: draw a 3D ROI, inspect the mean signal-intensity curve over time, save a BIDS-style derivative, and jump straight to the next case.

Loads BIDS-compliant 4D desc-hmc_DCE.nii[.gz] files, directories, or manifest lists
Live ROI curve preview with per-label and normalized views
BIDS-style derivative outputs with rater ID embedded in filenames
Auto-resumes at first unreviewed case; prefetches next image
Flag-and-skip for poor AIFs or missing baselines

GitHub

parametric_scripts

MATLAB

The ROCKETSHIP parametric mapping module, also usable as a stand-alone tool. Generates T1, T2, T2*, and ADC maps from multi-echo or inversion-recovery NIFTI series — a required pre-processing step for accurate DCE pharmacokinetic modeling.

T1 (inversion recovery), T2, T2*, and ADC map generation
GUI-based interface (fitting_gui)
Batch processing and parallel fitting support
NIFTI and DICOM input support

GitHub

Gpufit

CUDA / C++

A GPU-accelerated Levenberg–Marquardt curve-fitting library with Python and MATLAB wrappers. This PET/MRI Lab fork adds MRI-specific pharmacokinetic models and updated compiler/CUDA support. Used by ROCKETSHIP for fast voxel-wise model fitting.

Patlak, Tofts, Extended Tofts, Tissue Uptake, 2CXM, T1 FA Exponential
Full GPU (Gpufit) and CPU (Cpufit) parity for all MRI models
Pre-built binaries for Windows, Linux (CUDA 11.8–13.0), and macOS (CPU)
Python and MATLAB wrappers included
Based on Przybylski et al., Scientific Reports (2017)

GitHub Paper (Sci Rep 2017)

DCEPrep

Shell / Python

An end-to-end preprocessing and analysis pipeline for brain DCE-MRI. Wraps FSL, ANTs, FreeSurfer, ROCKETSHIP, and AutoAIF into a single configurable shell script. Handles VFA-based T1 mapping, bias field correction, z-axis normalization, head motion correction, AIF selection, Ktrans fitting, and automated QC reporting. A Docker image is provided for a consistent, reproducible environment.

BIDS-compliant input/output
VFA T1 mapping, bias field correction, z-axis normalization
Head motion correction via FSL mcflirt
AutoAIF integration for automatic AIF detection
Ktrans / vp mapping + per-case & population HTML QC reports
Docker image available for easy, reproducible deployment

GitHub

Which tool do I need?

Task	Tool
Convert raw scanner DICOMs into a BIDS dataset (first step before everything below)	dce2bids
Automated complete processing pipeline of BIDS DCE data (motion correction, alignment, artifact correction, create T1 maps, AIF detection, pharmacokinetic fitting, QC)	DCEPrep
Automated preprocessing only of BIDS DCE data (motion correction, alignment, artifact correction)	DCEPrep
Generate T1, T2, or ADC maps	parametric_scripts
AIF identification, automatically with deep learning (preferred)	AutoAIF
AIF identification, manually draw and save an AIF ROI (multi-rater)	AIFArtist
Scripted low level fast pharmacokinetic model fitting (useful to accelerate other DCE processing pipelines)	Gpufit
DCE processing with GUI	ROCKETSHIP
Visualize and compare fit quality and parametric maps	ROCKETSHIP (Module E)

Quick start

ROCKETSHIP

# Clone (includes submodules)
git clone --recursive \
  https://github.com/petmri/ROCKETSHIP.git

# In MATLAB, add ROCKETSHIP to the path, then:
run_parametric   % Step 1 — T1 maps
run_dce          % Step 2 — DCE fitting

AutoAIF

# Clone and set up environment
git clone https://github.com/petmri/AutoAIF.git
cd AutoAIF
python3 -m venv tf && source tf/bin/activate
pip install -r requirements.txt

# Run inference with pretrained weights
python main_vif.py --mode inference \
  --input_path /path/to/dce.nii.gz \
  --model_weight_path /path/to/weight.h5 \
  --save_output_path /path/to/output/

AIFArtist

# Clone and install dependencies
git clone https://github.com/petmri/AIFArtist.git
cd AIFArtist
pip install -r requirements.txt

# Launch the annotation app
python aif_artist.py /path/to/bids_dataset \
  --rater AB

Contact & contributing

Questions? Open an issue on the relevant repository or contact the lab maintainer at sabarnes@llu.edu. Contributions via pull request are welcome on all repositories.

View all repositories →