Modeling respiratory motion for reducing motion artifacts in 4D CT images

Area of Science:

• Medical Imaging
• Computational Anatomy
• Radiation Oncology

Background:

• Four-dimensional computed tomography (4D CT) is crucial for radiation treatment planning in thoracic and abdominal cancers.
• Image artifacts in 4D CT can compromise accurate tumor delineation and anatomical representation due to respiratory motion.
• Developing methods to reduce these artifacts is essential for effective cancer therapy.

Purpose of the Study:

• To present a patient-specific respiratory motion model for reducing artifacts in 4D CT images.
• To address image distortions caused by irregular breathing motion during 4D CT acquisition.
• To improve the accuracy of tumor delineation and anatomical shape representation in radiation treatment planning.

Main Methods:

• Deformable image registration was used to calculate displacement vector fields from 4D CT data.
• Principal Component Analysis (PCA) decomposed motion vectors into principal motion bases.
• A spline model parameterized projections to reconstruct displacement fields and synthesize artifact-reduced 4D CT images.

Main Results:

• Deformable registration significantly improved landmark accuracy compared to initial discrepancies.
• The synthesized 4D CT images closely matched original images, validating the motion model.
• Visual assessment confirmed substantial reduction in severe image artifacts and mitigated shape distortions in patient data.

Conclusions:

• A mathematical model effectively represents patient-specific respiratory motion from 4D CT data.
• The proposed approach successfully reduces irregular motion artifacts in 4D CT images.
• This method mitigates anatomical shape distortions caused by irregular breathing during 4D CT acquisition.