Assessing effects of low-dose CT attenuation correction on quantification and image quality on a LAFOV PET system with a modified NEMA-IQ phantom

Over the past decades, efforts have been directed toward increasing the sensitivity of PET systems to enable low-dose PET imaging and shorter scan times. The Biograph Vision Quadra PET/CT scanner (Siemens Healthineers, Knoxville, TN, USA) combines an increased solid angle coverage with time of flight, enabling a significant reduction in administered activity. To fully benefit from low-dose PET imaging, the dose of the CT scan should ideally also be reduced.

The CT radiation dose can be reduced by lowering the tube voltage and exposure as well as by the use of tin filtration. Reducing the CT dose must be carefully aligned with any impact that changing CT image quality and quantitative metrics may have on the PET quantification, for example via linear attenuation coefficients. We evaluate the impact of low-dose CT protocols on the PET quantification in soft-tissue-like and bone-equivalent materials using a customized phantom.

The phantom was assembled out of a NEMA-IQ phantom, several layers of saline and two bovine long bones, to simulate a patient positioned with arms down. A 10-min PET scan with 47 different accompanying CT scans was acquired. The CT tube voltage and exposure were varied between $70kV\mathrm{to}140kV$ , with and without tin filtration, and $7\mathrm{mAs}\mathrm{to}400\mathrm{mAs}$ , respectively. The effect of the different CT parameters on the PET image quality and quantitative accuracy was assessed via several quantitative metrics, including standardized uptake value (SUV) and signal-to-noise ratio, along with the CT dose length product (DLP).

The recorded DLP values suggest that using all tube voltages except $140kV$ combined with exposures in the range of $7mAs\mathrm{to}30mAs$ , can result in dose savings of $\ge$ $80\%$ compared to a typical clinical protocol ( $120kV$ , $50\mathrm{mAs}$ ). Compared to this typical protocol, HU differed more in bone ( $10\%\mathrm{to}40\%$ difference) than in soft tissue ( $1\%\mathrm{to}30\%$ difference), and more streak artefacts and photon starvation were seen when configuring low tube voltages and exposures, with an accompanying increase in CT noise. The observed artefacts were transferred with low intensity to the PET images via the attenuation map. The ${\mathrm{SUV}}_{\mathrm{mean}}$ remained within $2\%$ of the ${\mathrm{SUV}}_{\mathrm{mean}}$ of the typical clinical protocol for most tested protocols. The presence of artefacts in the CT images had a bigger impact on PET quantification in both soft tissue and bone ( ${\mathrm{SUV}}_{\mathrm{mean}}$ difference of $10\%\mathrm{to}20\%$ ) than any difference in HU. Protocols that resulted in artefacts showed an increased inaccuracy in attenuation correction and scatter estimation, as assessed via the "lung" residual error (up to $10\%\mathrm{to}20\%$ increase).

Even for low-dose CT protocols, PET quantification remained consistent in all NEMA-IQ spheres for many materials. Beam hardening and photon starvation in the CT image are likely to occur in specific body regions such as hips, shoulders or the abdomen for an arms-down scan and these effects had a bigger impact on the quantification than any difference in HU in both soft tissue and bone. This therefore should be considered when selecting suitable CT imaging parameters.