How does CT image noise affect 3D deformable image registration for image-guided radiotherapy planning?

Area of Science:

• Medical Imaging
• Radiation Oncology
• Image Processing

Background:

• Deformable image registration (DIR) is vital for transferring treatment planning data between CT scans.
• Cone-beam CT (CBCT) offers advantages but often exhibits higher image noise than fan-beam CT.
• Increased noise in CBCT may potentially compromise the accuracy of DIR and subsequent contouring.

Purpose of the Study:

• To quantify the sensitivity of deformable image registration to image noise.
• To evaluate the impact of simulated cone-beam CT noise on the accuracy of DIR.
• To compare registration accuracy between fan-beam CT and noisy CT images.

Main Methods:

• Utilized three CT studies per patient for five prostate cancer patients.
• Manual contours were drawn by three radiation oncologists on each CT.
• Simulated CBCT noise was added to target CTs, and registration accuracy was assessed by comparing transferred contours to manual contours.

Main Results:

• No significant degradation in registration accuracy was observed even with noise levels exceeding typical CBCT.
• The deformable registration process remained robust despite increased image noise.
• Changes in automatically transferred contours were monitored to detect registration accuracy shifts.

Conclusions:

• Noise levels in CBCT that could affect manual contouring do not reduce the accuracy of deformable image registration.
• Automated contour transfer via DIR is reliable even in the presence of significant image noise.
• DIR is a robust tool for radiation therapy planning, unaffected by CBCT-induced noise.