Variability of Radiomic Features in ¹⁸F-FDG PET/MR Brain Imaging: Impact of Scan Duration and Tracer Dose

Objectives: The purpose of this study is to optimize scan duration and tracer dose in radiomic analysis for <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup>F-FDG PET/MR brain imaging. Methods: Seventeen patients with brain cancer underwent PET/MR scans after injection of 3.70 MBq/kg <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup>F-FDG. PET images were split into 5 <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\text{PET}_{SD5}), 10(\text{PET}_{SD10}), 15(\text{PET}_{SD15})$</tex>, and 25 (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD25}$</tex>) min scan duration groups, and the injected dose for each group with 10% (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{D10\%}$</tex>), 20% (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{D20\%}$</tex>), 33% (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{D33\%}$</tex>) and 50% (<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{D50\%}$</tex>) dose levels were then simulated. 93 radiomic features were extracted and intraclass correlation coefficient (ICC) was calculated for each feature between groups, respectively. Results: ICC varied with the feature type, scan duration, and tracer dose. ICC of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD10} \text{PET}_{SD15}$</tex>, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD25}$</tex> in relation to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD5}$</tex> showed an obviously decrease with the lengthen of scan duration and the reduction of tracer dose. ICC on <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{D10\%}, \text{PET}_{D20\%}, \text{PET}_{D33\%}$</tex>, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{D50\%}$</tex> in relation to full-dose PET gradually increased with the enlargement of tracer dose and extended scan duration. ICC on different combinations with respect to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD5}$</tex> at full-dose presented similar trends in most features for <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD10}$</tex> at 50%, <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD15}$</tex> at 33%, and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD25}$</tex> at 20%. Besides, 28 features were non-sensitive to changes of scan durations and/or dose levels. Conclusions: <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{PET}_{SD10}$</tex> at 50% might be the clinically adaptable solution for <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">18</sup>F-FDG PET/MR brain imaging, showing reduced radiation burden, better compatibility and clinical feasibility. Changes in scan duration and/or tracer dose in PET should be approached with caution.