AI-based optimization for US-guided radiation therapy of the prostate

Area of Science:

• Medical Physics
• Radiation Oncology
• Robotics

Background:

• Volumetric ultrasound enables real-time intra-fractional target tracking in radiation therapy.
• Ultrasound probe placement can obstruct treatment beam directions, posing a planning challenge.

Purpose of the Study:

• To develop and evaluate an AI-based approach for optimizing ultrasound-robot pose and treatment beam selection.
• To assess the impact of robotic ultrasound guidance on radiation therapy planning for prostate cancer.

Main Methods:

• Combined CNN-based candidate beam selection with simulated annealing for robot setup.
• Employed linear optimization for treatment plan optimization.
• Studied 50 prostate cancer cases previously treated with CyberKnife.

Main Results:

• AI-based optimization significantly improved target coverage (93.66% to 97.20% average).
• Reduced total monitor units (MU) in some cases, especially for smaller targets.
• AI optimization yielded comparable plan quality with and without beam blocking considerations.

Conclusions:

• AI-based optimization provides a fast and effective method for robotic ultrasound-guided radiation therapy.
• The negative impact of ultrasound probe placement on plan quality is successfully minimized.