openrecon-ants

Screenshots from the MR Host

Siemens 7T Terra.X in XA60A.
Also tested on Siemens 3T Cima.X on XA61(-SP01)

Brainmasking, Debias, Denoise

Sequence: Non-selective 3D SPACE.
From left to right: Original, brain mask (SynthStrip), N4BiasFieldCorrection (ANTs) in brain mask, DenoiseImage (ANTs) in brain mask after N4BiasFieldCorrection.
From top to bottom: 3D FLAIR with Circular Polarization (CP), 3D FLAIR with Universal Pulses (UP), 3D DIR with UP.

ANTs using ANTsPy in OpenRecon.
Brain masking is performed by SynthStrip

SkullStripping

Sequence: Angio TOF
From left to right: Original, SkullStripped using brainmask from Synthstrip, N4BiasFieldCorrection, DenoiseImage
From top to bottom: SAG MIP, COR MIP, TRA MIP

Features

This OR performs ANTs image operations :

• N4BiasFieldCorrection
• DenoiseImage
• N4BiasFieldCorrection then DenoiseImage (default)
• DenoiseImage then N4BiasFieldCorrection
• None (only for brain masking / skull stripping)

Brain mask usage :

• Apply ANTs in brainmask (keep outside the mask intact)
• Skull stripping then ANTs (keep only in mask)
• None (only ANTs, no masking)

There is an option, a checkbox, to save original images and intermediate images. (default is True)

Based on https://github.com/benoitberanger/openrecon-template

Build

Requirements for building :

• python 3.12
• jsonschema

Python environment manager is strongly recomanded :

conda create --name openrecon-ants
conda install python=3.12
pip install jsonschema

Build with :

python build.py

Offline test and dev

Python modules :

• ismrmrd
• pydicom
• pynetdicom
• antspyx
• torch
• surfa

pip install ismrmrd pydicom pynetdicom antspyx torch surfa

Follow guidelines in https://github.com/benoitberanger/openrecon-template

TODO

• Add fields in the UI to tune N4BiasFieldCorrection and DenoiseImage
• Add fields in the UI to tune the mask (erode ? delate ? ...)