Model-Based Iterative Reconstruction

Model-based iterative reconstruction (MBIR) sets a new direction in CT image quality with virtually noise-free images and industry-leading low-contrast resolution. This improvement in low-contrast resolution is a breakthrough made possible in iterative reconstruction built on a knowledge-based model. In addition to these image quality benefits, for the first time, physicians are also able to simultaneously combine these improvements with lower doses.

To address the emerging challenges in CT reconstruction, CT manufacturers more recently released an IR technology called MBIR. Similar to standard iterative construction, MBIR also belongs to the category of statistical IR methods, which are also denoted as MBIR. As with other implementations of statistical IR methods, MBIR is characterized by (a) the use of statistical weighting in the raw-data space followed by a back projection (unfiltered or filtered), (b) the application of a regularization function consisting of a smoothness constraint and a statistical model in the image space, and (c) the use of forward projection (i.e., data re-projection) with an adequate CT system model. The latter forward-projection step generates pseudo raw data’ that are compared to the measured raw data. The process of repeatedly comparing the measured raw data with pseudo raw data contributes primarily to the cancellation of spiral artifacts and, to a lesser extent, to noise reduction.

MBIR is the first knowledge-based solution that overcomes the motion-sensitivity associated with traditional model-based solutions; allowing it to be used in even the most advanced acquisitions, such as cardiac CTA. Enabled by next-generation supercomputing hardware and reconstruction algorithm innovation, its reconstruction speed allows MBIR to be used in even the most time demanding applications, such as trauma.

Key benefits

Industry-leading low-contrast resolution; Significantly lower dose while simultaneously improving image quality; Virtually noise-free images; Significantly improved low-contrast detectability; Significantly improved spatial resolution; First knowledge-based iterative reconstruction for gated acquisitions; Fast reconstruction; and Integrated design with minimal siting impact.