GPU-based iterative cone-beam CT reconstruction using tight frame regularization.

Area of Science:

• Medical Imaging
• Computational Imaging
• Radiation Oncology

Background:

• Serial cone-beam computed tomography (CBCT) scans in image-guided radiation therapy pose a clinical concern due to cumulative x-ray imaging dose.
• Reducing radiation exposure while maintaining diagnostic image quality is crucial for patient safety and treatment efficacy.

Purpose of the Study:

• To develop a fast graphic processing unit (GPU)-based algorithm for reconstructing high-quality CBCT images from undersampled and noisy projection data.
• To lower the imaging dose in image-guided radiation therapy by enabling dose reduction strategies through efficient reconstruction.

Main Methods:

• Developed an iterative tight-frame (TF)-based CBCT reconstruction algorithm incorporating sparsity constraints under a TF basis.
• Employed a multi-grid method for computational acceleration, implemented on a GPU for high efficiency.
• Validated the algorithm using digital (NCAT) and physical (Catphan) phantoms, and a clinical head-and-neck patient case.

Main Results:

• Achieved high computational efficiency, reconstructing a 512 × 512 × 70 CBCT image in approximately 5 minutes.
• Demonstrated successful reconstruction of CBCT images from undersampled and low mAs projection data.
• Quantitatively confirmed high image quality using modulation-transfer function and contrast-to-noise ratio analyses, outperforming state-of-the-art TV algorithms in certain aspects.

Conclusions:

• The developed TF-based GPU algorithm effectively reconstructs high-quality CBCT images from low-dose, undersampled data.
• This approach offers a promising solution for reducing radiation exposure in image-guided radiation therapy.
• The algorithm shows potential for clinical implementation, improving patient safety without compromising image fidelity.