How can spherical-CNNs improve ML-based diffusion MRI parameter estimation?

Tobias Goodwin-Allcock, Jason McEwen, Robert Gray, Parashkev Nachev, Gary Zhang

Keywords: Spherical-CNN, Spherical CNN, diffusion MRI, dMRI, DWI, diffusion tensor, DT, fractional anisotropy, FA

TL;DR: Demonstrated advantages of spherical CNNs over conventional fully connected networks at estimating microstructure indices

Abstract:

This paper demonstrates spherical convolutional neural networks (spherical-CNN) offer distinct advantages over conventional fully connected networks (FCN) at estimating rotation-invariant indices of tissue microstructure from diffusion MRI (dMRI). Such microstructure indices are valuable for identifying pathology and quantifying its extent. However, current clinical practise commonly acquires dMRI data consisting of only 6 diffusion weighted images (DWIs), limiting the accuracy and precision of estimated microstructure indices.
Machine learning (ML) has been proposed to address this challenge. However, existing ML-based methods are not robust to differing dMRI sampling schemes, nor are they rotationally equivariant. Lack of robustness to sampling schemes requires a new network to be trained for each scheme, complicating the analysis of data from multiple sources. The lack of rotational equivariance potentially prevents these methods from estimating the same microstructure viewed from different angles consistently. Here, we show spherical-CNNs represent a compelling alternative that is robust to new gradient schemes as well as offering rotational equivariance. We show the latter can be leveraged to decrease the number of training datapoints required.

Registration: I acknowledge that publication of this at MIDL and in the proceedings 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: methodological development

Primary Subject Area: Image Acquisition and Reconstruction

Secondary Subject Area: Learning with Noisy Labels and Limited Data

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.