Deep learning-based segmentation of high-resolution computed tomography image data outperforms commonly used automatic bone segmentation methods

July 04, 2021

Daniella M. Patton (1), Emilie N. Henning (2), Rob W. Goulet (1), Sean K. Carroll (1), Erin M.R. Bigelow (1), Benjamin Provencher (3), Nicolas Piché (3), Mike Marsh (3), Karl J. Jepsen (1), Todd L. Bredbenner (2)
bioRxiv, July 2021. DOI: 10.1101/2021.07.27.453890


Bone segmentation; threshold; deep learning; CNNs


Segmenting bone from background is required to quantify bone architecture in computed tomography (CT) image data. A deep learning approach using convolutional neural networks (CNN) is a promising alternative method for automatic segmentation. The study objectives were to evaluate the performance of CNNs in automatic segmentation of human vertebral body (micro-CT) and femoral neck (nano-CT) data and to investigate the performance of CNNs to segment data across scanners. Scans of human L1 vertebral bodies (microCT [North Star Imaging], n=28, 53µm3) and femoral necks (nano-CT [GE], n=28, 27µm3 ) were used for evaluation.

How Our Software Was Used

Dragonfly was used for image segmentation and for the training of two-dimensional U-Net CNNs.

Author Affiliation

(1) Department of Orthopaedic Research, University of Michigan, Ann Arbor, MI 48109
(2) Mechanical and Aerospace Engineering Department, University of Colorado Colorado Springs, Colorado Springs, CO, 80918.
(3) Object Research Systems, Montreal, QC, CA.