Iterative reconstruction for dual energy CT with an average image-induced nonlocal means regularization

Area of Science:

• Medical Imaging
• Computed Tomography
• Image Reconstruction

Background:

• Reducing radiation dose in dual-energy computed tomography (DECT) is crucial but can increase image noise.
• Filtered backprojection (FBP) reconstruction of low-dose DECT data often results in excessive noise, impacting diagnostic certainty.

Purpose of the Study:

• To develop a novel image reconstruction scheme for low-dose DECT acquisitions.
• To improve the quality of DECT images and material decomposition by reducing noise and artifacts.

Main Methods:

• Incorporation of an averaged image induced non-local means (aviNLM) regularization into the penalized weighted least-squares (PWLS) framework.
• The PWLS-aviNLM scheme utilizes DECT data-dependent variance and considers image similarity across energy levels.
• Validation on digital, physical phantoms, and patient data.

Main Results:

• The PWLS-aviNLM algorithm quantitatively outperformed FBP, PWLS-TV, and PWLS-NLM algorithms.
• Qualitative results showed superior image quality with finer details and fewer noise-induced artifacts.
• Demonstrated effectiveness in improving DECT reconstruction and material decomposition.

Conclusions:

• The proposed PWLS-aviNLM algorithm is effective for reconstructing high-quality DECT images from low-dose acquisitions.
• This method enhances diagnostic confidence by minimizing noise and artifacts.
• The study highlights the potential of aviNLM regularization for advanced DECT imaging.