The R&D team, led jointly by GE's Bruno De Man and Ge Wang of Rensselaer Polytechnic Institute, conducted an extensive study underwritten by a grant from the National Institutes of Health (NIH) to evaluate various approaches to CT system design, including variations on conventional and unconventional architectures, with the hope that they might find an approach with the potential to outperform conventional architectures. The team described the comparative evaluation methods that they used, so that other researchers who wish to analyse proposed CT architectures can utilize these as a reference framework.

The team hypothesized that a dedicated cardiac CT scanner would require a 300-mm field-of-view (FOV), 160-mm Z coverage, 50-mm temporal resolution, 20-lp/cm spatial resolution, and sufficient power to maintain image noise below 10 Hounsfield Units (HU). Images need to be acquired with the lowest radiation dose possible to produce diagnostic image quality without degrading artefacts. They evaluated six proposed CT architectures, including systems with one, two and three beam lines (source–detector pairs) to address the challenge of temporal resolution.

The researchers analysed cost- and performance-related criteria for the comparative evaluation, each with specific detailed methods for scoring. These included:
• The inherent Radon completeness of the raw data with respect to being close to 100% to avoid cone-beam artefacts.
• Integrated effective power – the total capability to deliver X-ray energy to a patient within a given scan time, thus determining the lower limit of noise in raw data.
• The capability to provide required temporal resolution through high-speed data acquisition without including the effect of motion-correction algorithms.
• Dose efficiency – the ability to deliver dose uniformly throughout the volume-of-interest and to measure the transmitted X-rays with minimal loss of information.
• The costs to develop new hardware technology, new software algorithms that would not have residual artefacts, and actual production costs.

As an example, the susceptibility to cone-beam artefacts was analysed to estimate the fundamental challenges associated with image reconstruction for each of the proposed CT architectures. To evaluate scatter, a critical concern in wide-cone CT architecture, the team first quantified the reduction in dose efficiency due to scatter and then evaluated forward scatter and cross-scatter for multiple sources irradiating a volume simultaneously. They also analysed the impact of an antiscatter grid, utilization efficiency of the projection ray redundancy, virtual bowtie capability and geometric irradiation efficiency.

The team then created a detailed prioritization matrix for system architectures for a single-, dual- and triple-source CT (SSCT, DSCT and TSCT), as well as a dual twin-Z CT (DTZCT), a triple inverse-geometry CT (TIGCT) and a ring source, dual-rotating-detector CT (OSDRT). They describe the architectures and the performance of each proposed CT system in detail in their article.

The researchers determined that the architectures that offered the best cost–performance were systems based on traditional X-ray tubes using one, two or three beam lines. They eliminated the proposed TIGCT and OSDRT architectures with distributed sources due to cost and related performance characteristics. As modelled, the DTZCT architecture did not offer any significant improvements that could justify its increased cost.

Performance relative to the SSCT increased with the additional beam lines in DSCT and TSCT, but at additional cost. "We found that CT system architectures based on the third-generation geometry provide nearly linear improvement versus the increased cost of additional beam lines, although similar performance improvement could be achieved with advanced motion-correction algorithms. The third-generation architectures outperform even the most promising of the proposed architectures that deviate substantially from the traditional CT system architectures," the authors wrote. They concluded by saying that their detailed evaluation confirms the validity of the current trend in commercial CT scanner design.

Related stories

• Ultrahigh-resolution CT raises the bar
• Neural networks reduce noise in low-dose CT
• CTA protocol slashes radiation dose
• Omni-tomography: the ultimate in multimodality imaging?