Related Experiment Videos
Validation of fully automatic brain SPET to MR co-registration
1Department of Nuclear Medicine, The Queen Elizabeth Hospital, Woodville, SA, Australia.
European Journal of Nuclear Medicine
|February 7, 2001
Summary
This study validates automatic co-registration of single-photon emission tomography (SPET) to magnetic resonance (MR) brain imaging. The best method achieved an intrinsic registration error of 1.2 mm, comparable to PET-MR accuracy.
Area of Science:
- Medical Imaging
- Neuroscience
- Radiology
Background:
- Fully automatic co-registration of functional to anatomical brain images is validated for PET-MR but not SPET-MR.
- Accurate co-registration is crucial for integrating multi-modal brain imaging data.
Purpose of the Study:
- To evaluate five fully automatic methods for technetium-99m hexamethylpropylene amine oxime (99mTc-HMPAO) SPET to MR co-registration.
- To assess the accuracy and reliability of these automated registration techniques.
Main Methods:
- Utilized six fiducial markers attached to the skin, visible in both SPET and MR.
- Determined marker locations to sub-pixel precision and calculated registration errors (FRE, SRE, IRE).
- Excluded subjects with excessive fiducial marker distortion in SPET or MR.
Main Results:
- The best method achieved a mean intrinsic registration error (IRE) of 1.2 mm (max 3.3 mm), surface registration error (SRE) of 1.5 mm (max 4.3 mm), and fiducial registration error (FRE) of 2.4 mm (max 5.0 mm).
- All evaluated methods demonstrated a mean IRE below 3 mm.
- The accuracy of the most accurate method was comparable to published PET-MR co-registration results.
Conclusions:
- Fully automatic SPET to MR co-registration is feasible and accurate.
- With proper calibration, intra-subject cerebral SPET to MR registration accuracy under 2 mm is achievable.
- These findings support the clinical utility of automated SPET-MR image fusion.