Development of an Integrated PET-CT Simulation Pipeline for Virtual Imaging Trials and Biomedical Imaging Research

Abstract

Positron emission tomography-computed tomography (PET-CT) is an essential imaging technique that merges anatomical and physiological data, aiding in lesion detection, tumor localization, and treatment planning. Enhancing PET-CT protocols is key to advancing clinical applications and emerging areas such as theranostics and personalized dosimetry. Conducting physical imaging trials is constrained by ethical, financial, and logistical limitations. Virtual imaging trials provide a solution by offering a structured approach to evaluating imaging systems, refining protocols, and analyzing how acquisition parameters impact image quality in a controlled setting. Although PET and CT simulators exist separately, PET simulations often depend on idealized attenuation maps rather than CT-based attenuation correction, limiting their ability to account for variations in CT acquisition. The development of a fully integrated PET-CT simulation framework would help bridge this gap, creating a controlled platform for assessing imaging protocols, lesion detectability, and quantitative accuracy.This study aimed to develop and validate an integrated PET-CT simulation pipeline that ensures input and geometric concordance between PET and CT simulations. By incorporating simulated CT attenuation correction maps instead of assuming idealized attenuation, this pipeline enables a more comprehensive assessment of PET-CT acquisition and reconstruction protocols. The goal was to create a flexible and validated platform capable of simulating clinically relevant scenarios, including respiratory motion, scanner-specific imaging variations, lesion detectability in multi-modality imaging, and protocol optimizations. The integrated PET-CT simulation pipeline was developed by combining two validated simulation platforms: SimSET for PET emission modeling and DukeSim for CT simulations. A unified integer-based voxelized computational phantom framework ensured input consistency across both modalities, while a standardized coordinate system established geometric alignment. The pipeline was verified and validated using both a NEMA IEC hot sphere phantom and a NEMA spatial resolution test to assess PET system resolution accuracy and compare simulated PET-CT data with clinical PET-CT scans. Key quantitative metrics—including contrast recovery coefficients (CRC), standard uptake values (SUV), and activity concentrations—were evaluated to assess image fidelity. To further demonstrate its utility, pilot virtual imaging trials were conducted using XCAT phantoms in clinically relevant scenarios. The first set of virtual trials investigated lesion detectability using dual-modality imaging. Simulations modeled a thoracic lesion with high attenuation contrast to its surroundings and a brain lesion with similar attenuation characteristics but high metabolic activity. These trials demonstrated how PET-CT imaging enhances lesion delineation in both cases, emphasizing the complementary nature of PET and CT. Additionally, a study assessing the effects of CT respiratory motion on PET quantification was performed using an XCAT phantom with an 18F-FDG biodistribution modeled from clinical data. PET reconstructions were performed using different CT attenuation correction maps derived from varying CT acquisition protocols, including inspiration breath-hold, expiration breath-hold, and free-breathing motion with different respiratory phase binning strategies. The integrated pipeline successfully demonstrated geometric concordance, with a maximum PET-CT centroid difference of 1.83 mm, well within the clinically acceptable range and below the typical PET voxel size of 3–4 mm. The NEMA spatial resolution tests confirmed strong agreement with clinical PET spatial resolution measurements, with minor deviations in radial and tangential full-width-at-half-maximum (FWHM) and full-width-at-tenth-maximum (FWTM) values at larger radial distances. PET reconstructions using DukeSim-generated CT attenuation maps more closely resembled clinical PET-CT images compared to reconstructions using ideal attenuation maps, reinforcing the necessity of realistic CT simulations for attenuation correction. The pilot virtual imaging trials demonstrated the impact of PET-CT integration on lesion detectability, with CT aiding tumor boundary delineation in high-contrast scenarios and PET improving detectability in cases where attenuation contrast was minimal. The respiratory motion study confirmed that CT acquisition variability significantly influences PET SUV measurements, particularly in the liver, myocardium, and blood pool, with variations in attenuation correction protocols leading to quantifiable differences in PET quantification. This study presents a novel PET-CT simulation pipeline that integrates realistic CT attenuation correction, addressing a key limitation of existing PET simulation platforms that rely on idealized attenuation maps. By incorporating patient-specific CT variability, the pipeline enables controlled investigations of PET-CT acquisition protocols, lesion characterization, and quantification accuracy, providing a valuable tool for virtual imaging trials, theranostics, radiomics, and personalized dosimetry applications. The ability to simulate respiratory motion effects, spatial resolution characteristics, lesion detectability, and scanner-specific imaging variations further expands its utility for optimizing multi-modality imaging protocols. Future work will focus on validating the pipeline across additional PET-CT scanner models, expanding its use in theranostics dosimetry, and conducting large-scale virtual trials to refine PET-CT imaging techniques. Additionally, the platform’s ability to systematically assess the interplay between PET and CT for lesion characterization and imaging protocol optimization will continue to support advancements in precision medicine and quantitative imaging research.