Model-Based Iterative Reconstruction for Advanced Baggage Screening
F3-A5 (Phase 1)

Download the 2013 Project Report

While traditional direct reconstruction algorithms such as filtered back projection (FBP) and direct Fourier method (DFM) depend on the analytic inversion of sinogram data, model-based reconstruction methods rely on iterative optimization of a statistical model of both the acquired data and the unknown image. Model-based iterative reconstruction (MBIR) potentially offers many important advantages over traditional methods of direct reconstruction for the security screening of checked baggage. It has the potential to reduce metal artifacts, improve resolution, reduce artifacts such as cupping that can systematically distort CT number estimates, incorporate tighter integration of reconstruction and segmentation, and support reconstruction of scanners with a small number of projections. All these improvements have the potential to improve the detection/false alarm tradeoff for CT security screening systems.

The objective of this research is to implement an MBIR algorithm on a widely deployed multi-slice helical CT security scanner, and assess qualitatively and quantitatively the improvements over direct reconstruction. The MBIR implementation entails the accurate modeling of both the system geometry, including subtle manufacturing deviations, as well as the photon and electronic noise characteristics. The quality assessment was carried out using a set of 12 vendor-provided bag scans, selected with the guidance of the ALERT Center. The results demonstrate significant quality improvements over the native DFM reconstructions, including improved metal artifact suppression, spatial resolution, and CT value uniformity.

 

We have shown that an alternative method of image reconstruction provides the sort of quality improvement, on data from a currently certified scanner, that is likely to move the detection/false alarm probability curve above its present placement.
F3-A5 Relevance & Transition: ALERT Year 5 Annual Report
Project Leader
  • Charles A. Bouman
    Professor
    Purdue University
    Email

Faculty and Staff Currently Involved in Project
  • Ken D.Sauer
    Professor
    Notre Dame University
    Email

Students Currently Involved in Project
  • Pengchong Jin
    Purdue University
  • Eri Haneda
    Purdue University