Semiautomated segmentation of the human spine based on echoplanar images

Giovanni Giulietti, Paul E. Summers, Diana Ferraro, Carlo A. Porro, Bruno Maraviglia, Federico Giove

Research output: Contribution to journalArticlepeer-review

Abstract

The number of functional magnetic resonance imaging (fMRI) studies performed on the human spinal cord (SC) has considerably increased in recent years. The lack of a validated processing pipeline is, however, a significant obstacle to the spread of SC fMRI. One component likely to be involved in any such pipeline is the process of SC masking, analogous to brain extraction in cerebral fMRI. In general, SC masking has been performed manually, with the incumbent costs of being very time consuming and operator dependent. To overcome these drawbacks, we have developed a tailored semiautomatic method for segmenting echoplanar images (EPI) of human spine that is able to identify the spinal canal and the SC. The method exploits both temporal and spatial features of the EPI series and was tested and optimized on EPI images of cervical spine acquired at 3 T. The dependence of algorithm performance on the degree of EPI image distortion was assessed by computing the displacement warping field that best matched the EPI to the corresponding high-resolution T 2 images. Segmentation accuracy was above 80%, a significant improvement over values obtained with similar approaches, but not exploiting temporal information. Geometric distortion was found to explain about 50% of the variance of algorithm classification efficiency.

Original languageEnglish
Pages (from-to)1429-1436
Number of pages8
JournalMagnetic Resonance Imaging
Volume29
Issue number10
DOIs
Publication statusPublished - Dec 2011

Keywords

  • EPI
  • FMRI
  • K-means
  • Segmentation
  • Spinal cord
  • Temporal standard deviation

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging
  • Biomedical Engineering

Fingerprint Dive into the research topics of 'Semiautomated segmentation of the human spine based on echoplanar images'. Together they form a unique fingerprint.

Cite this