DIANA - Detectability Investgations using Artificial Nodal Additions
Reimund Bayerlein, M. Xia, H. Xie, B. A. Spencer, Jinsong Ouyang, Georges El Fakhri, Lorenzo Nardo, C. Liu, Ramsey Derek Badawi · 2024
Positron Emission Tomography (PET) imaging often encounters challenges in quantifying lesion detectability and contrast recovery due to the absence of known in vivo ground truths. To address this, we introduce DIANA (Detectability Investigations using Artificial Nodal Additions), a Monte Carlo-based tool designed to improve lesion detectability analysis. DIANA permits the insertion of artificial lesions into human subject list-mode data with precise activity concentrations, thereby enabling a detailed assessment of image quality metrics. In this study, we demonstrated DIANA by analyzing the influence of image smoothing and the number of iterations on image contrast and the contrast-to-noise ratio (CNR) in PET scans. The tool was tested using data from the uEXPLORER total-body PET/CT scanner, with artificial lesions between 4 mm and 8 mm in diameter introduced into the scans of a 73-year-old female. Images were reconstructed using an in-house OSEM algorithm, and lesion detectability was quantified through the lesion-to-background ratio (LBR) and the contrast recovery coefficient (CRC) across different count densities and frame lengths. Results demonstrated a dependency of CRC on frame length and lesion size, with notable degradation at shorter frame lengths. The study also explored the effects of Gaussian smoothing on contrast to noise ratio (CNR), highlighting the expected trade-off between lesion visibility and image noise. Overall, DIANA proved effective in validating PET imaging techniques and enhancing the quantitative accuracy of lesion detectability, paving the way for further applications in varied clinical scenarios and dynamic imaging. Future work will extend the utility of DIANA by incorporating patient motion and diverse body mass indices to reflect more realistic clinical conditions.