Nuklearmedizin 2025; 64(01): 77-78
DOI: 10.1055/s-0045-1804372
Abstracts │ NuklearMedizin 2025
Wissenschaftliche Vorträge
Onkologie – Bildgebung und Therapie

Kinetic Parameter Estimation from Dynamic PET with Optimised Acquisition Protocol of LAFOV PET Scanners

Authors

  • X He

    1   Chair for Computer Aided Medical Procedures, Technical University of Munich, Garching b. München, Deutschland

  • F De Benetti

    1   Chair for Computer Aided Medical Procedures, Technical University of Munich, Garching b. München, Deutschland

  • K Tehlan

    1   Chair for Computer Aided Medical Procedures, Technical University of Munich, Garching b. München, Deutschland
    2   Department of Radiology, University Hospital Augsburg, Augsburg, Deutschland

  • T Wendler

    1   Chair for Computer Aided Medical Procedures, Technical University of Munich, Garching b. München, Deutschland
    2   Department of Radiology, University Hospital Augsburg, Augsburg, Deutschland
    3   Institute of Digital Medicine, University Hospital Augsburg, Neusäß, Deutschland
Further Information
Also available at
 
 

Ziel/Aim: Kinetic parameters (KP) derived from Dynamic PET (dPET) analysis with compartment models are clinically valuable. However, conventional methods for voxel-wise KP computation are slow. De Benetti et al. [1] showed that it is possible to use deep learning to generate KP maps (KPMs) which are consistent with the KP at organ level previously published in the literature [2], while reducing the computation time and taking into consideration the interactions of neighboring voxels. In this work, we use the approach [1] to find optimal acquisition protocols for DynamicPET able to either limit the total acquisition time or use lower temporal resolution while generating satisfactory KPMs.

Methodik/Methods: The dataset consists of 23 patients with various oncological indications undergoing LAFOV 18F-FDG DynamicPET acquisition of 65 min, including the image-derived input function (IDIF) [3]. We compared the KPMs generated from the full acquisition (65 min) [1] with those predicted using only the initial 45 min and those using double duration frames. The mean percentage errors (MPE) per organ was used as the evaluation metric.

Ergebnisse/Results: When using only the initial 45 min, the MPE in the computation of the KP is<15%. When using double duration frames, if the frames are longer near the IDIF peak, the MPE is ~ 20% whereas if the frames are longer far from the IDIF peak (i.e. at least 30 sec away), it is<15%.

Schlussfolgerungen/Conclusions: We showed that the generation of KPMs is possible using shorter acquisition (45 instead of 65 min) in 18F-FDG dPET, thus increasing the patient throughput. Moreover, we found that the use of longer frames (e.g. 4 instead of 2 sec or 20 instead of 10 sec) is feasible, but must be done carefully because frames with longer durations around the IDIF peak can lead to large errors. A similar experiment was performed by Karakatsanis [4] using another method, but led to the same results. Overall we showed that the generation of KPMs is not limited to last generation LAFOV PET scanners making the use of KPMs in diagnosis more widespread.



Publication History

Article published online:
12 March 2025

© 2025. Thieme. All rights reserved.

Georg Thieme Verlag KG
Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany