Intra-operative Localization of Brachytherapy Implants Using Intensity-based Registration.

Area of Science:

• Medical Imaging
• Radiation Oncology
• Image Registration

Background:

• Prostate brachytherapy requires accurate visualization of both the prostate boundary and implanted radioactive seeds.
• Transrectal ultrasound (TRUS) visualizes the prostate boundary but struggles with seed detection.
• Fluoroscopy clearly shows implanted seeds but lacks prostate boundary definition.

Purpose of the Study:

• To develop and validate an intensity-based registration method for combining TRUS and fluoroscopy data.
• To achieve accurate intra-operative dosimetry in prostate brachytherapy.

Main Methods:

• Proposed an automated, intensity-based registration technique using mutual information.
• Filtered and compounded TRUS images, then registered them to a fluoroscopy-derived seed implant model.
• Evaluated performance using a phantom study with 48 implanted seeds.

Main Results:

• Achieved an average seed registration error of 0.78 mm relative to ground truth.
• Demonstrated robustness to false positives (error < 1.01 mm at 31% rate) and false negatives (error < 0.96 mm at 31% rate).
• Optimized TRUS filtering using Bayesian combination of adaptive thresholding, phase congruency, and beam profile compensation.

Conclusions:

• The automated registration method shows excellent accuracy and robustness in phantom studies.
• This technique holds promise for clinically adequate performance in human prostate brachytherapy.
• Accurate seed localization via combined imaging enhances intra-operative dosimetry.