Whole-body PET/CT imaging: combining software- and hardware-based co-registration

Area of Science:

• Medical Imaging
• Radiology
• Image Processing

Background:

• Combined PET/CT offers superior co-registration but clinical routine often introduces local misalignments.
• Misaligned CT data for attenuation correction (CT-AC) can bias reconstructed PET tracer distribution.
• Improving PET/CT quality in routine clinical practice is essential.

Purpose of the Study:

• To investigate the feasibility of retrospective co-registration techniques for aligning CT and PET images prior to CT-AC.
• To enhance the quality of combined PET/CT imaging in clinical settings.
• To evaluate automated non-rigid registration algorithms for whole-body PET/CT.

Main Methods:

• Commercial software was used for rigid and non-rigid CT-PET image alignment.
• Co-registration accuracy was assessed visually and via PET tracer uptake patterns.
• A prototype ITK implementation of the B-spline algorithm was evaluated for non-rigid registration.

Main Results:

• Manual rigid co-registration led to unacceptable misalignment, especially in peripheral whole-body image areas.
• Manual non-rigid co-registration showed minor local distortions but was largely successful.
• Automated rigid and non-rigid methods using Mutual Information failed to co-register images.
• The automated non-rigid B-spline co-registration technique demonstrated promising results with MR-CT data.

Conclusions:

• Retrospective registration of CT and uncorrected PET images can improve AC-PET image quality in clinical settings.
• Currently, no validated, clinically viable commercial registration software is in routine use.
• An automated, validated, non-rigid co-registration algorithm is needed for whole-body PET/CT.
• The presented B-spline approach shows suitability for retrospective WB-PET/CT alignment.