Development of a prediction model for target positioning by using diaphragm waveforms extracted from CBCT projection...

Area of Science:

• Medical Imaging
• Radiation Oncology
• Image-guided Therapy

Background:

• Accurate tumor targeting is crucial for effective lung cancer radiotherapy.
• Respiratory motion significantly impacts target localization accuracy in Cone-Beam Computed Tomography (CBCT).
• Current methods often rely on implanted markers, which carry risks.

Purpose of the Study:

• To develop and evaluate a prediction model (PM) for target positioning.
• Utilize diaphragm waveforms extracted from CBCT projection images for motion prediction.
• Assess the efficacy of Amsterdam Shroud (AS) signal-based correction for improved accuracy.

Main Methods:

• Nineteen lung cancer patients underwent orthogonal rotational kV x-ray imaging.
• Developed four regression-based prediction models (PM_CL, PM_4D-CBCT, PM_AS-4D-CBCT) using surrogate and target motion data.
• Compared prediction errors using different waveform extraction and correction techniques, including the AS signal.

Main Results:

• The 4D-CBCT-based model (PM_4D-CBCT) improved prediction accuracy by 12.8% over the clinical model (PM_CL).
• Amsterdam Shroud (AS) signal-based correction further enhanced accuracy, achieving 84.0%-85.3% prediction within 3 mm.
• Statistically significant improvements were observed with AS-corrected models (PM_AS-4D-CBCT) compared to PM_4D-CBCT.

Conclusions:

• The AS signal-based correction significantly improves target position prediction accuracy from CBCT images.
• The developed PM_AS-4D-CBCT demonstrates the efficacy of using diaphragm waveforms for motion prediction.
• This approach offers a promising method for predicting target positions without requiring implanted gold markers.