Use of the CT component of PET-CT to improve PET-MR registration: demonstration in soft-tissue sarcoma

Phys Med Biol. 2007 Dec 7;52(23):6991-7006. doi: 10.1088/0031-9155/52/23/015. Epub 2007 Nov 12.

Abstract

We have investigated improvements to PET-MR image registration offered by PET-CT scanning. Ten subjects with suspected soft-tissue sarcomas were scanned with an in-line PET-CT and a clinical MR scanner. PET to CT, CT to MR and PET to MR image registrations were performed using a rigid-body external marker technique and rigid and non-rigid voxel-similarity algorithms. PET-MR registration was also performed using transformations derived from the registration of CT to MR. The external marker technique gave fiducial registration errors of 2.1 mm, 5.1 mm and 5.3 mm for PET-CT, PET-MR and CT-MR registration. Target registration errors were 3.9 mm, 9.0 mm and 9.3 mm, respectively. Voxel-based algorithms were evaluated by measuring the distance between corresponding fiducials after registration. Registration errors of 6.4 mm, 14.5 mm and 9.5 mm, respectively, for PET-CT, PET-MR and CT-MR were observed for rigid-body registration while non-rigid registration gave errors of 6.8 mm, 16.3 mm and 7.6 mm for the same modality combinations. The application of rigid and non-rigid CT to MR transformations to accompanying PET data gives significantly reduced PET-MR errors of 10.0 mm and 8.5 mm, respectively. Visual comparison by two independent observers confirmed the improvement over direct PET-MR registration. We conclude that PET-MR registration can be more accurately and reliably achieved using the hybrid technique described than through direct rigid-body registration of PET to MR.

Publication types

  • Evaluation Study
  • Validation Study

MeSH terms

  • Algorithms*
  • Humans
  • Image Enhancement / methods*
  • Image Interpretation, Computer-Assisted / methods*
  • Magnetic Resonance Imaging / methods*
  • Positron-Emission Tomography / methods*
  • Reproducibility of Results
  • Sarcoma / diagnosis*
  • Sensitivity and Specificity
  • Soft Tissue Neoplasms / diagnosis*
  • Subtraction Technique*
  • Tomography, X-Ray Computed / methods*