SE(3)-Equivariant and Noise-Invariant 3D Rigid Motion Tracking in Brain MRI

Rigid motion tracking is paramount in many medical imaging applications where movements need to be detected, corrected, or accounted for. Modern strategies rely on convolutional neural networks (CNN) and pose this problem as rigid registration. Yet, CNNs do not exploit natural symmetries in this tas...

Full description

Saved in:
Bibliographic Details
Published inIEEE transactions on medical imaging Vol. 43; no. 11; pp. 4029 - 4040
Main Authors Billot, Benjamin, Dey, Neel, Moyer, Daniel, Hoffmann, Malte, Abaci Turk, Esra, Gagoski, Borjan, Ellen Grant, P., Golland, Polina
Format Journal Article
LanguageEnglish
Published United States IEEE 01.11.2024
Subjects
Adult
Algorithms
Biomedical imaging
Brain - diagnostic imaging
Convolution
Feature extraction
Female
fetal MRI
Humans
Imaging, Three-Dimensional - methods
Magnetic resonance imaging
Magnetic Resonance Imaging - methods
motion tracking
Movement - physiology
Neural Networks, Computer
rigid registration
SE-equivariant CNNs
Three-dimensional displays
Tracking
Transforms
Online AccessGet full text
ISSN0278-0062
1558-254X
1558-254X
DOI10.1109/TMI.2024.3411989

Cover

More Information
Summary:Rigid motion tracking is paramount in many medical imaging applications where movements need to be detected, corrected, or accounted for. Modern strategies rely on convolutional neural networks (CNN) and pose this problem as rigid registration. Yet, CNNs do not exploit natural symmetries in this task, as they are equivariant to translations (their outputs shift with their inputs) but not to rotations. Here we propose EquiTrack, the first method that uses recent steerable SE(3)-equivariant CNNs (E-CNN) for motion tracking. While steerable E-CNNs can extract corresponding features across different poses, testing them on noisy medical images reveals that they do not have enough learning capacity to learn noise invariance. Thus, we introduce a hybrid architecture that pairs a denoiser with an E-CNN to decouple the processing of anatomically irrelevant intensity features from the extraction of equivariant spatial features. Rigid transforms are then estimated in closed-form. EquiTrack outperforms state-of-the-art learning and optimisation methods for motion tracking in adult brain MRI and fetal MRI time series. Our code is available at https://github.com/BBillot/EquiTrack .
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0278-0062
1558-254X
1558-254X
DOI:10.1109/TMI.2024.3411989