Quantitative characterization of metastatic disease in the spine. Part I. Semiautomated segmentation using atlas-based deformable registration and the level set method

Med Phys. 2007 Aug;34(8):3127-34. doi: 10.1118/1.2746498.

Abstract

Quantitative assessment of metastatic disease in bone is often considered immeasurable and, as such, patients with skeletal metastases are often excluded from clinical trials. In order to effectively quantify the impact of metastatic tumor involvement in the spine, accurate segmentation of the vertebra is required. Manual segmentation can be accurate but involves extensive and time-consuming user interaction. Potential solutions to automating segmentation of metastatically involved vertebrae are demons deformable image registration and level set methods. The purpose of this study was to develop a semiautomated method to accurately segment tumor-bearing vertebrae using the aforementioned techniques. By maintaining morphology of an atlas, the demons-level set composite algorithm was able to accurately differentiate between trans-cortical tumors and surrounding soft tissue of identical intensity. The algorithm successfully segmented both the vertebral body and trabecular centrum of tumor-involved and healthy vertebrae. This work validates our approach as equivalent in accuracy to an experienced user.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Algorithms
  • Automation
  • Bone Neoplasms / diagnosis*
  • Bone Neoplasms / diagnostic imaging*
  • Bone Neoplasms / pathology
  • Breast Neoplasms / diagnostic imaging*
  • Breast Neoplasms / pathology
  • Female
  • Humans
  • Image Processing, Computer-Assisted
  • Middle Aged
  • Neoplasm Metastasis / diagnosis*
  • Neoplasm Metastasis / diagnostic imaging*
  • Observer Variation
  • Spinal Neoplasms / diagnosis*
  • Spinal Neoplasms / diagnostic imaging*
  • Spine / diagnostic imaging*
  • Spine / pathology
  • Tomography, X-Ray Computed / methods