Scoliosis Measurement on DXA Scans Using a Combined Deep Learning and Spinal Geometry Approach

Emmanuelle Bourigault, Amir Jamaludin, Timor Kadir, Andrew Zisserman

Keywords: Scoliosis measurement, image segmentation, deep learning, computational geometry

TL;DR: We propose improvements to an automated method for scoliosis measurement using a spline geometric representation of the spine and pseudo-labelling the segmentation for domain adaptation.

Abstract: We propose improvements to an automated method for scoliosis measurement. Our main novelty is the use of a spline to better model the curve of the spine, and we employ pseudo- labelling to re-train the segmentation step to mitigate the domain gap when adapting to a new dataset. We obtain promising results with a good fit of our smoothed curve to approximate the spinal midpoints in severe scoliosis cases, and obtain good agreement against human ground-truth. This work is relevant for improving the severity grading of scoliosis and potentially aiding in the treatment management of scoliosis.

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: recently published or submitted journal contributions

Primary Subject Area: Segmentation

Secondary Subject Area: Application: Radiology

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.