Complexity and accuracy of image registration methods in SPECT-guided radiation therapy

Area of Science:

• Medical Imaging
• Radiotherapy
• Image Registration

Background:

• Accurate co-registration of functional imaging to CT scans is crucial for radiotherapy (RT) planning.
• Functional imaging, like SPECT, provides valuable data for precise RT targeting.

Purpose of the Study:

• To evaluate six image registration methods for SPECT-guided radiotherapy treatment planning.
• To assess the accuracy and suitability of different registration techniques, including rigid and non-rigid methods.

Main Methods:

• Six image registration methods were evaluated, ranging from 3D affine transforms to non-rigid diffeomorphic demons and level set algorithms.
• Perfusion SPECT scans from ten lung cancer patients were registered to planning CT scans.
• Registration accuracy was assessed using intensity differences and target registration error.

Main Results:

• Non-rigid registration methods demonstrated higher accuracy compared to rigid methods.
• However, some non-rigid methods produced deformation fields causing unacceptable SPECT intensity changes.
• Intensity histogram differences were largest for diffeomorphic demons and level set methods.

Conclusions:

• Intensity-based validation alone is insufficient for SPECT/CT registration in radiotherapy treatment planning (RTTP).
• Proper evaluation of image registration necessitates the use of multiple accuracy metrics.
• Careful consideration of deformation field integrity is vital when using non-rigid registration for SPECT/CT in RT planning.