Real-time CBCT reconstructions using Krylov solvers in repeated scanning procedures

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Real-time cone-beam computed tomography (CBCT) is crucial for image-guided surgery.
• Limited measurements in CBCT necessitate efficient reconstruction methods.
• Prior reconstructions can aid in subsequent scans with local changes.

Purpose of the Study:

• Introduce an efficient iterative solver for real-time CBCT reconstruction.
• Utilize Prior Image Constrained Compressed Sensing (PICCS) and Krylov subspace methods.
• Improve artifact reduction and reconstruction speed in dynamic imaging scenarios.

Main Methods:

• Developed a novel iterative solver incorporating PICCS regularization.
• Leveraged Krylov subspace methods for computational efficiency.
• Applied the method to sequences of under-sampled CBCT scans with local object changes.

Main Results:

• Demonstrated effectiveness in both synthetic and real CT data.
• PICCS framework significantly reduced image artifacts.
• The new solver exhibited superior speed compared to existing alternatives.

Conclusions:

• The proposed iterative solver offers an efficient and effective solution for real-time CBCT reconstruction.
• PICCS with Krylov subspace methods is highly beneficial for dynamic imaging scenarios.
• This method holds promise for improving image-guided surgery and other time-sensitive applications.