Iterative CT reconstruction via minimizing adaptively reweighted total variation.

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• Iterative reconstruction using total variation (TV) minimization is effective for accurate CT imaging from under-sampled projections.
• Reducing projection data further leads to over-smoothing artifacts, particularly at structure boundaries.

Purpose of the Study:

• To develop a practical algorithm for enhancing TV-minimization based CT reconstruction using very limited projection data.
• To address the limitations of current methods in significantly reducing projection views without compromising image quality.

Main Methods:

• Proposed an L-0 norm approach, theoretically superior for reducing projection views, approximated via an adaptive weighting scheme.
• Implemented a series of TV minimizations where weights adapt based on image gradients from preceding iterations.
• Reconstruction iteration ceases upon observing minimal differences in weighted TV values between consecutive images.

Main Results:

• Evaluated on digital and physical phantoms, demonstrating significant improvements in CT reconstruction.
• Achieved over a 5-fold reduction in reconstruction errors compared to conventional TV minimization using only 20 equiangular projections.
• Showcased enhanced spatial resolution in the reconstructed images.

Conclusions:

• Adaptive reweighting of TV in iterative CT reconstruction effectively reduces the required number of projection views.
• The proposed method achieves comparable or superior image quality with substantially fewer projections.
• This advancement offers a practical solution for low-dose or rapid CT imaging scenarios.