We demonstrate an object tracking method for 3D images with fixed computational cost and state-of-the-art performance. Previous methods predicted transformation parameters from convolutional layers. We instead propose an architecture that does not include either flattening of convolutional features or fully connected layers, but instead relies on equivariant filters to preserve transformations between inputs and outputs (e.g. rot./trans. of inputs rotate/translate outputs). The transformation is then derived in closed form from the outputs of the filters. This method is useful for applications requiring low latency, such as real-time tracking. We demonstrate our model on synthetically augmented adult brain MRI, as well as fetal brain MRI, which is the intended use-case.
@article{arxiv.2103.10255,
title = {Equivariant Filters for Efficient Tracking in 3D Imaging},
author = {Daniel Moyer and Esra Abaci Turk and P Ellen Grant and William M. Wells and Polina Golland},
journal= {arXiv preprint arXiv:2103.10255},
year = {2021}
}