Deep learning-based Fast Volumetric Image Generation for Image-guided Proton Radiotherapy

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Artificial Intelligence in Medicine

Background:

• Image-guided radiation therapy (IGRT) precision is crucial for effective cancer treatment.
• Fast imaging techniques are needed to improve IGRT, especially for lung cancer patients requiring gating.
• FLASH radiotherapy offers potential benefits for organ-at-risk sparing without compromising tumor control.

Purpose of the Study:

• To develop and validate a deep learning (DL)-based framework for rapid volumetric image reconstruction.
• To enable accurate target localization for lung cancer patients undergoing image-guided radiation therapy.
• To evaluate the framework's performance in the context of proton FLASH therapy.

Main Methods:

• A four-module framework was developed: kV x-ray projection acquisition, DL-based volumetric image generation, image quality analysis, and proton water equivalent thickness (WET) evaluation.
• Volumetric images were reconstructed using kV projection pairs from four different source angles.
• Thirty lung cancer patient datasets with 4D CT scans were utilized for evaluation.

Main Results:

• The optimal kV projection source angles for volumetric image reconstruction were identified as 135° and 225°.
• The framework achieved patient-averaged performance metrics including mean absolute error of 75±22 HU, peak signal-to-noise ratio of 19±3.7 dB, structural similarity index measure of 0.938±0.044, and WET error of -1.3%±4.1%.
• The developed framework demonstrated rapid volumetric image delivery capabilities.

Conclusions:

• The proposed DL-based framework enables fast volumetric image reconstruction for precise target localization in lung cancer patients.
• This technology has the potential to guide proton FLASH treatment delivery systems, enhancing therapeutic precision and safety.
• The identified optimal kV projection angles contribute to improved image quality for advanced radiotherapy applications.