National Center for Image Guided Therapy

• NCIGT Wiki

• About Us
  • Overview
  • Research Labs
  • Research Cores
  • Research Projects
  • DBPs
  • People
• Resources
  • Publication DB
  • Image Gallery
  • Downloads
  • AMIGO
  • News and Events
  • Contact Us

EPI Distortion Correction

EPI distortion. Figure 1 shows an example of EPI distortion. The red edges illustrate structure from a T1-weighted structural image. EPI image intensities appear in gray scale. The EPI brain is seen to extend out of the brain region of the structural MRI and the ventricles are mis-registered.

Computation Core researchers within the NCIGT have been investigating the issue of distortions in echo-planar (EPI) MRI that, caused by magnetic susceptibility artifacts, limit the accuracy by which fMRI activations can be situated in the anatomy of subjects and patients. The limitation is concerning to neurosurgical clinicians who use preoperative fMRI to perform functional mapping of the brain for language, for example. During prolonged fMRI sessions subject motion has continually become problematic, particularly in regards to neurosurgical patients.

The most successful approach to minimizing EPI distortions makes use of acquired magnetic "field maps" that can be used for subsequent dewarping of the associated EPI images (a capability that is included in the FSL and SPM software packages).

Improved agreement of dewarped EPI and structural MRI is seen in the area of the ventricles. In Figure 2, the image on the left was corrected using FUGUE, while the image on the right was corrected using a synthetic predicted field map.

Subject motion during prolonged fMRI sessions can be an issue in fMRI scanning, particularly with neurosurgical patients. The data processing pipelines for fMRI activation detection routinely perform "motion correction" using a registration algorithm to compensate for the motion, however, the current field-map-based approach to dewarping EPI is not robust to subject motion: the acquired field map is no longer valid if the subject's head has moved.

NCIGT researchers developed a new approach to EPI dewarping that uses synthetic field maps that are predicted from subject structural MRI by a process of segmentation, registration, and applied physical calculation. A potential advantage of this approach is that the synthetic field map can be rapidly regenerated at any new orientation of the head, making the method robust to subject motion. In preliminary testing, the approach has produced results that are visually comparable to a standard field-map based method (PRELUDE and FUGUE) and quantitatively better than a similar approach that predicts field maps using a CT-based atlas.

Back to Research Projects.

Involved Investigators

• William "Sandy" Wells PhD, NCIGT Core PI
• Alexandra Golby, MD, NCIGT Core PI
• Kinh Tieu, PhD
• Clare Poynton, PhD Candidate