SU-E-J-41: Fluoroscopy Based Adaptive Setup Approach for Thoracic Cancer IGRT.

Area of Science:

• Medical Imaging
• Machine Learning
• Radiation Oncology

Background:

• Accurate patient positioning is crucial in thoracic cancer image-guided radiation therapy (IGRT).
• Respiratory motion during treatment can lead to significant setup deviations, impacting treatment efficacy and normal tissue toxicity.
• Real-time adaptive adjustments require precise quantification of motion and positional variations.

Purpose of the Study:

• To develop a fluoroscopy imaging-based approach for quantifying respiratory motion magnitude and phase variations.
• To determine online patient setup deviations in thoracic cancer IGRT.
• To enable real-time adaptive patient position adjustments.

Main Methods:

• A numerical phantom with simulated respiratory motion (varying frequency, amplitude, offset, phase shift) was used.
• White noise (SNR=5) was added to simulate clinical image quality.
• A manifold-based machine learning technique constructed a respiratory motion model, and MAP solutions quantified phase and position shifts.

Main Results:

• The approach successfully detected variations in motion patterns between image sets.
• It demonstrated insensitivity to frequency changes and image noise up to SNR=5.
• The method effectively captured and quantified changes in motion amplitude, position shift, and phase shift.

Conclusions:

• An effective mathematical approach quantifies motion differences between pre-treatment and planning images using machine learning on fluoroscopy.
• Application during online patient setup allows separation and adjustment of positioning deviations from respiratory motion.
• This facilitates minimized normal tissue toxicity in gated IGRT.