Megavoltage CT imaging quality improvement on TomoTherapy via tensor framelet

Area of Science:

• Medical Physics
• Image Reconstruction
• Computational Imaging

Background:

• TomoTherapy megavoltage (MV) CT imaging is crucial for radiotherapy guidance.
• Improving MV CT image quality, especially under low-dose or undersampled conditions, remains a challenge.

Purpose of the Study:

• To investigate the feasibility of a novel tensor framelet reconstruction technique for enhancing TomoTherapy MV imaging quality.
• To evaluate the technique's performance using full-view and partial-view data.

Main Methods:

• Image reconstruction formulated as a least-squares L1-type optimization problem using tensor framelets for regularization.
• High-order image derivatives regularized in L1 norm across multiple levels and directions.
• Efficient solution using the Split Bregman method and a GPU-based parallel algorithm.
• Comparison with filtered backprojection (FBP) and total variation (TV) methods on phantoms and patients.

Main Results:

• The tensor framelet method demonstrated superior image quality compared to FBP and TV methods.
• The technique proved robust with as little as 25% of projection views.
• GPU-based reconstruction of a 40-slice, 1mm-resolution 3D image took approximately 2 minutes.

Conclusions:

• A GPU-based tensor framelet reconstruction method effectively improves TomoTherapy MV CT image quality.
• The method shows significant potential for low-dose imaging, enabling up to a 75% reduction in projection views while maintaining high quality.