Keywords: 7T Cardiac MRI, segmentation, synthesis, style transfer, bias field simulation
TL;DR: By using cardiac images at 3T to create synthetic images at 7T, we are able to more accurately segment 7T cardiac image
Abstract: To train a field strength agnostic cardiac segmentation network, we propose two novel augmentation techniques that allow us to transform 3T images to synthetic 7T images: by i) simulating $B_1$ distribution to approximate the 7T bias field and ii) style transfer using an unpaired 3T-to-7T GAN model. Data augmentation with these two methods improved the average Dice score over all classes by 22% and 25% respectively, on our 7T test dataset. Furthermore, the average performance on a 1.5T and 3T dataset were maintained.
Registration: I acknowledge that acceptance of this work at MIDL requires at least one of the authors to register and present the work during the conference.
Authorship: I confirm that I am the author of this work and that it has not been submitted to another publication before.
Paper Type: novel methodological ideas without extensive validation
Primary Subject Area: Segmentation
Secondary Subject Area: Image Synthesis
Confidentiality And Author Instructions: I read the call for papers and author instructions. I acknowledge that exceeding the page limit and/or altering the latex template can result in desk rejection.