Geometry-Aware Attenuation Learning for Sparse-View CBCT Reconstruction.

Area of Science:

• Medical Imaging
• Computer Vision
• Radiology

Background:

• Cone Beam Computed Tomography (CBCT) is crucial for clinical imaging but requires high radiation doses due to numerous 2D X-ray projections.
• Sparse-view CBCT reconstruction aims to reduce radiation exposure, but existing deep learning and neural rendering methods have limitations in quality and efficiency.

Purpose of the Study:

• To develop a novel, efficient, and accurate method for sparse-view CBCT reconstruction.
• To address the limitations of current deep learning and neural rendering approaches in terms of image quality and computational time.

Main Methods:

• Introduced a geometry-aware encoder-decoder framework for 3D CBCT image reconstruction.
• Utilized a 2D CNN encoder to extract multi-view 2D features from sparse X-ray projections.
• Employed a geometry-guided back-projection to create a 3D volumetric feature map.
• Applied a 3D CNN decoder for final 3D CBCT image recovery, respecting projection geometry.

Main Results:

• Achieved exceptional reconstruction quality even with extremely sparse views (e.g., 5 or 10 projections).
• Demonstrated significant time efficiency compared to existing methods.
• Validated the framework's adaptability without individual training for sparse-view scenarios.

Conclusions:

• The proposed geometry-aware framework offers a promising solution for low-dose sparse-view CBCT reconstruction.
• The method effectively balances image quality and computational efficiency.
• This approach has the potential to reduce patient radiation exposure in clinical practice.