SVRTK/aide-svrtk

SVRTK packaged as a MONAI Application Package (MAP), including for deployment on AIDE

aide-svrtk

AI-driven, automated SVR reconstruction packaged as an AIDE Application, based on the open-source MONAI Application Package (MAP) standard.

View repo · Report Bug · Request Feature

Overview

The slice-to-volume reconstruction toolkit (SVRTK) is an image-based registration framework for reconstruction of 3D volumes from multiple 2D image slices. SVRTK is used extensively for performing motion-corrected reconstruction of MRI data.

Currently, this MAP implements SVRTK for reconstruction of 2D fetal brain MRI data into motion-corrected 3D volumes. Further MAPs for reconstruction of other fetal organs are in development and coming soon.

AIDE is an open-source platform for the deployment of AI applications in healthcare settings. This repo, created and led by Tom Roberts, packages automated 3D fetal brain MRI reconstruction into a MONAI Application Package (MAP) for execution on AIDE, and other platforms compatible with the MAP standard.

aide-svrtk MAP workflow

The input to the aide-svrtk MAP is multiple 2D DICOM Series, each containing multi-slice 2D MRI data. The output is a 3D MRI DICOM Series contained the motion-corrected reconstruction.

The aide-svrtk MAP consists of three operators:

1. dcm2nii_operator.py – converts the input 2D MRI DICOM Series into NIfTI format required by SVRTK, using dcm2niix
2. fetal_mri_3d_brain_recon_operator.py – runs automated fetal brain SVRTK reconstruction, in three main steps: a. AI-driven brain masking b. Slice-to-volume registration reconstruction c. AI-driven brain reorientation
3. nii2dcm_operator.py – converts the SVRTK output NIfTI into a 3D DICOM Series, using nii2dcm

Developers

Prerequisites

• GPU-enabled machine
• Docker
  • Required for running the MAP

Setup

1. Download

git clone https://github.com/SVRTK/aide-svrtk.git

2. Setup virtual env

cd aide-svrtk

python -m venv venv
source venv/bin/activate

pip install --upgrade pip setuptools wheel
pip install -r requirements.txt

3. Create input and output directories

mkdir input output

Run MAP source code with MONAI Deploy

1. Ensure Python venv running
2. Copy DICOM Series files to input/ directory
   • Standard or Enhanced DICOMs

# input - 2D multi-stack DICOM Series files
# output – 3D SVRTK-reconstructed DICOM Series
monai-deploy exec app -i input/ -o -output/

Build and run as MONAI Application Package (MAP)

1. Ensure Python venv running
2. Ensure Docker running
3. Copy DICOM Series files within input/ directory

Important: we create an initial MAP map-init upon which we build any 3rd-party non-Python software (e.g. dcm2niix). The final MAP is called map

# Initial packaging of MAP
monai-deploy package app -t ghcr.io/svrtk/aide-svrtk/map-init:0.2.0 -r requirements.txt -l DEBUG

# Build 3rd-party software on top of MAP
docker build -t ghcr.io/svrtk/aide-svrtk/map:0.2.0 app/

# Test MAP with MONAI Deploy
monai-deploy run ghcr.io/svrtk/aide-svrtk/map:0.2.0 input/ output/

# Push initial MAP and final MAP to GHCR
docker push ghcr.io/svrtk/aide-svrtk/map-init:0.2.0
docker push ghcr.io/svrtk/aide-svrtk/map:0.2.0

Optional

Enter Docker container for testing

docker run --gpus all -it --rm -v local/path/to/aide-svrtk/input:/var/monai/input/ --entrypoint /bin/bash ghcr.io/svrtk/aide-svrtk/map:0.2.0

Run on specified GPU if machine has >1 available

CUDA_VISIBLE_DEVICES=2 monai-deploy run ghcr.io/svrtk/aide-svrtk/map:0.2.0 input/ output/

Running on AIDE

To run aide-svrtk MAP on AIDE, two files from the app/workflows directory are required, namely:

• fetal-3d-brain-mri.json – AIDE Clinical Workflow file.
  • Effectively a sequence of tasks run on AIDE
• fetal-brain-3d-recon-argo-template.yaml – Argo Workflow Template file.
  • Called by the AIDE Clinical Workflow file. Executes the aide-svrtk MAP using Argo. This is the central task within the AIDE Clinical Workflow file.