An iterative tomosynthesis reconstruction using total variation combined with non-local means filtering.

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Compressed sensing (CS) theory has advanced reconstruction from sparse data, improving artifact reduction.
• Iterative reconstruction methods, incorporating local and non-local regularization, are crucial for limited view angle imaging.

Purpose of the Study:

• To introduce a novel 3D iterative image reconstruction method, (ART+TV)NLM.
• To enhance image quality in tomosynthesis by addressing noise and out-of-focus blur.

Main Methods:

• Developed a hybrid method combining Algebraic Reconstruction Technique (ART) with Total Variation (TV) minimization and Non-Local Means (NLM) filtering.
• Applied TV minimization for edge-preserving noise reduction.
• Utilized NLM filtering to suppress slice blur by exploiting image structure similarity.

Main Results:

• Simulations using a tomosynthesis system and 3D phantom demonstrated superior performance of (ART+TV)NLM.
• Visual inspection revealed significant improvements in image quality compared to ART and ART+TV methods.
• Quantitative metrics like RMSE, SSIM, and SNR showed enhanced results.

Conclusions:

• The proposed (ART+TV)NLM method achieves significant improvements in convergence rate and image quality.
• Proper selection of NLM parameters is key to maximizing these benefits.
• This technique offers a promising solution for high-quality tomosynthesis reconstruction.