Benjamin Billot, Neel Dey, Daniel Moyer, Malte Hoffmann, Esra Abaci Turk, Borjan Gagoski, P. Ellen Grant, & Polina Golland. (2024). SE(3)-Equivariant and Noise-Invariant 3D Rigid Motion Tracking in Brain MRI. IEEE Transactions on Medical Imaging, 1-12. https://doi.org/10.1109/TMI.2024.3411989
Tracking movement accurately in medical imaging, like MRI scans of the brain, is crucial to get clear and useful images. Traditional methods use a type of artificial intelligence called convolutional neural networks (CNNs) to detect and correct these movements, but CNNs aren’t good at handling rotations of the images.
The researchers have developed a new method called EquiTrack, which uses advanced CNNs that can handle both shifts and rotations in the images. However, these advanced CNNs struggle with noisy images, which are common in medical imaging. To fix this, the researchers combined their CNNs with a noise-reduction tool to separate irrelevant noise from important image features.
EquiTrack was tested on brain MRI images from both adults and fetuses and was found to perform better than current leading methods. This new technique promises more accurate tracking of movement, leading to clearer and more reliable MRI scans. The code for EquiTrack is freely available online for other researchers to use and build upon. Source: https://github.com/BBillot/EquiTrack
