Registration accuracy for MR images of the prostate using a subvolume based registration protocol

Area of Science:

• Medical Imaging
• Radiotherapy
• Image Registration

Background:

• Magnetic resonance imaging (MRI) offers superior soft tissue contrast for radiotherapy workflows compared to computed tomography.
• Image registration is crucial for patient positioning and assessing therapy response using repeated imaging.
• This study investigates the impact of registration volume size on accuracy for MR image-based prostate radiotherapy.

Purpose of the Study:

• To determine the relationship between registration accuracy and registration volume size in MR image-based prostate radiotherapy.
• To optimize subvolume-based rigid registration protocols for enhanced precision in prostate imaging.
• To reduce uncertainty in prostate positioning for improved radiotherapy accuracy.

Main Methods:

• Ten patients underwent four T2-weighted MRI scans during radiotherapy.
• Rigid image registration utilized a mean square distance metric and step gradient optimizer with varying registration volumes.
• Registration precision was assessed by the center of mass distance of manually defined prostates; optimal volume minimized standard deviation.

Main Results:

• Full volume registration showed the highest uncertainty in prostate position, particularly in the anterior-posterior (AP) direction.
• Optimizing the registration volume to the prostate gland improved the standard deviation of the center of mass distance by 3.9 mm (p < 0.001) in the AP direction.
• The optimal registration volume size was found to be 0 mm margin around the prostate gland.

Conclusions:

• High-precision tissue registration is essential for both radiotherapy setup and response assessment using repeated prostate MRI.
• Subvolume-based registration significantly reduces prostate registration uncertainty to approximately 1 mm (1 SD).
• This contrasts with the several millimeters of uncertainty observed with whole-pelvis registration, highlighting the benefit of optimized subvolume approaches.