Volumetric tumor tracking from a single cone-beam X-ray projection image enabled by deep learning

Area of Science:

• Medical Physics
• Oncology
• Artificial Intelligence

Background:

• Radiotherapy accuracy is limited by respiratory motion affecting tumor position.
• Accurate tumor tracking is crucial for effective radiation delivery.
• Current methods may involve higher radiation doses or complex setups.

Purpose of the Study:

• To develop a deep learning-based method for accurate 3D tumor tracking using single-angle X-ray images.
• To enable precise volumetric tumor localization during radiotherapy.
• To reduce radiation dose while maintaining high localization accuracy.

Main Methods:

• Utilized a deep learning model for volumetric tumor tracking from single-angle X-ray projections.
• Aligned intraoperative 2D X-rays with pre-treatment 3D CT scans for tumor localization and segmentation.
• Developed a patient-specific model using data augmentation, style correction, and registration networks.
• Validated the method on patient lung data and phantoms.

Main Results:

• Achieved high localization precision for 3D tumor positioning.
• Demonstrated effectiveness at reduced radiation doses.
• Successfully tracked tumors despite respiratory-induced motion.

Conclusions:

• The deep learning-anchored volumetric tumor tracking method enhances radiotherapy precision.
• This approach offers accurate tumor localization with lower radiation exposure.
• It represents a significant advancement in image-guided radiation therapy.