Metal artifact reduction in computed tomography using local models in an image block-iterative scheme

Area of Science:

• Medical Imaging
• Computational Science
• Materials Science

Background:

• Metal artifacts pose a significant challenge in medical imaging, degrading image quality and potentially leading to misdiagnosis.
• Iterative reconstruction techniques offer improved artifact reduction but often increase computational demands.

Purpose of the Study:

• To develop and evaluate a computationally efficient iterative reconstruction method for reducing metal artifacts.
• To test the hypothesis that complex reconstruction models are only necessary in metal-vicinity regions.

Main Methods:

• A novel method automatically segments reconstruction volumes into metal and nonmetal regions (patches).
• Different reconstruction models (IMPACT for metal, MLTRC for nonmetal) are applied locally to these patches.
• Sequential updating of patches is employed, mimicking a grouped coordinate algorithm.

Main Results:

• The proposed local model approach (IMPACT+MLTRC) achieves image quality comparable to using complex models throughout.
• This method significantly reduces streak artifacts and improves the convergence of metal regions.
• Computational complexity is reduced by a factor of approximately 2.6 compared to full complex model reconstruction.

Conclusions:

• Complex reconstruction models can be applied locally to metal regions for effective artifact reduction without compromising image quality.
• This localized approach offers a substantial computational advantage over global application of complex models.
• The method shows promise for improved reconstruction of structures near or within metal implants compared to traditional methods like projection completion.