Abstract
241962
Introduction: In PET/CT imaging, CT transmission data are used to correct PET emission data for attenuation and scatter. Typically, current PET/CT systems calculate the linear attenuation coefficient (μ) for 511-keV photons from CT Hounsfield units. However, a significant limitation of CT-based PET attenuation correction (AC) is the introduction of artifacts due to misalignment between PET and CT images, leading to challenges in clinical interpretation and quantitative inaccuracies. In long axial field-of-view (LAFOV) PET/CT systems, such as the uMI Panorama GS with a 148-cm field, a full-length attenuation CT is necessary to fully utilize the scanner's sensitivity and perform AC, yet it often proves redundant for diagnostic purposes. Consequently, there is a growing need for a CT-less AC solution in LAFOV PET/CT. This study explores a CT-less AC solution developed for the uMI Panorama GS, which features a 148-cm PET/CT system with sub-200-ps time-of-flight (TOF) and sub-3-mm spatial resolution, aiming to eliminate the dependence on CT images.
Methods: The CT-less AC consists of two technical innovations: (1) A convolutional neural network (CNN) is employed to generate an attenuation map (μ-DL) from non-attenuation-corrected (NAC) PET images. The network inputs these NAC PET images and outputs matched CT attenuation maps (μ-CT). The network was trained using PET/CT data previously acquired from uMI Panorama and uEXPLORER scanners. (2) To enhance AC accuracy, a bed template is incorporated into the μ-DL. An empty bed plate is pre-scanned as the bed template which covers the entire scan range. Owing to the network output only covering the human body parts, the μ-DL lacks bed sheets and other clinical accessories. The clinical study was approved by the Independent Ethics Committee of the First People's Hospital of Kunshan (2023-02-027-K01, 2023-03-055-K01).
Results: Figure 1 displays the results of μ-CT, μ-DL, CT-AC PET, CT-less AC PET, and their difference maps. The coronal views reveal similarities between μ-CT and μ-DL, and only minor differences between CT-AC and CT-less AC OSEM results. Lung nodules exhibit comparable SUVMAX values in both CT-AC and CT-less AC PET images. The transverse views, aligned with the blue dotted line in the coronal view, show that μ-DL can distinguish between water and air in the stomach, similar to μ-CT, a feature not as evident in the PET images. Figure 2 presents OSEM results for three patients at different dose levels using both CT-AC and CT-less AC. Figure 2A features a normal-dose patient with a high BMI (35.6), demonstrating that CT-less AC provides effective AC even in high BMI subjects. The red arrow highlights the reduction of PET/CT mismatch artifacts with CT-less AC. Figure 2B shows a half-dose patient who underwent a two-bed scan, again with minimal discrepancy between CT-AC and CT-less AC, including reduced banana artifacts at the liver-lung boundary in the difference image. Figure 2C depicts a low-dose clinical case (approximately 1/6 of normal dose) with over 2 hours post-injection time, illustrating that CT-less AC maintains comparable performance to CT-AC.
Conclusions: The developed CT-less AC technique effectively addresses PET-CT misalignment and reduces artifacts introduced by CT in the AC process. It proves the feasibility of performing AC without relying on CT images. The application of this method to 18F-FDG scans on the uMI Panorama GS demonstrates that CT-less AC can provide quantification results comparable to traditional CT AC, while reducing motion artifacts. Additionally, its applicability in low-dose PET studies is noteworthy. The adoption of CT-less imaging enhances the flexibility of LAFOV PET/CT protocols and shows great promise for clinical applications. This is particularly relevant in pediatric studies, dual-tracer or multi-tracer investigations, and longitudinal studies where managing cumulative radiation dosage is crucial.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.