Bone marrow segmentation based on a combined consideration of transverse relaxation processes and Dixon oscillations. Academic Article uri icon

Overview

abstract

  • The aim of this study was to demonstrate that gradient-echo sampling of single spin echoes can be used to isolate the signal from trabecular bone marrow, with high-quality segmentation and surface reconstructions resulting from the application of simple post-processing strategies. Theoretical expressions of the time-domain single-spin-echo signal were used to simulate signals from bone marrow, non-bone fatty deposits and muscle. These simulations were compared with and used to interpret signals obtained by the application of the gradient-echo sampling of a spin-echo sequence to image the knee and surrounding tissues at 1.5 T. Trabecular bone marrow has a much higher reversible transverse relaxation rate than surrounding non-bone fatty deposits and other musculoskeletal tissues. This observation, combined with a choice of gradient-echo spacing that accentuates Dixon-type oscillations from chemical-shift interference effects, enabled the isolation of bone marrow signal from surrounding tissues through the use of simple image subtraction and thresholding. Three-dimensional renderings of the marrow surface were then readily generated with this approach - renderings that may prove useful for bone morphology assessment, e.g. for the measurement of femoral anteversion. In conclusion, understanding the behavior of signals from bone marrow and surrounding tissue as a function of time through a spin echo facilitates the segmentation and reconstruction of bone marrow surfaces using straightforward post-processing strategies that are typically available on modern radiology workstations.

publication date

  • February 11, 2016

Research

keywords

  • Bone Marrow
  • Image Processing, Computer-Assisted
  • Magnetic Resonance Imaging

Identity

Scopus Document Identifier

  • 84969331965

Digital Object Identifier (DOI)

  • 10.1002/nbm.3498

PubMed ID

  • 26866627

Additional Document Info

volume

  • 29

issue

  • 5