AI based segmentation for dosimetry
Manual versus artificial intelligence-based segmentations as a pre-processing step in whole-body dosimetry calculations (conference abstract)
- 123
Publications
Take a look
Manual vs AI based segmentation for dosimetry
Manual versus artificial intelligence-based segmentations as a pre-processing step in whole-body dosimetry calculations
by Joyce van Sluis, Walter Noordzij, Lars Edenbrandt, Elisabeth G. E. de Vries, Adrienne H. Brouwers, and Ronald Boellaard
Poster presentation at the EANM 2021 conference
Abstract
Aim/Introduction
Over the last decades, labelling of monoclonal antibodies (MAbs) with zirconium-89 (89Zr) allowed whole body assessment of MAb distribution and tumour targeting over time with molecular imaging. The main advantage of 89Zr is the long half-life of 78.4 h matching the pharmacokinetic behaviour of antibodies, making it suitable for labelling of MAbs.
The long physical half-life of 89Zr and the long biological half-life of MAbs may cause high radiation burden and/or limits the amount of activity that can be administered, which in turn limits image quality. It is therefore important to obtain reliable radiation dose estimates to optimize the amount of activity that can be administered while keeping radiation burden within acceptable limits.
Organ segmentation is required for whole-body dosimetry but is a very time-consuming task. Therefore, we explored the possibility of using an AI based automated segmentation tool as a pre-processing step for calculating the organ and whole-body effective doses.
Materials and Methods
Retrospective PET/CT data of six patients undergoing treatment with 89Zr-labelled pembrolizumab were included in this study. Manual organ segmentations were performed using in-house developed software and biodistribution information was obtained. Using the activity biodistribution information, residence times were calculated. The obtained residence times served as input for OLINDA XLM version 1.0 (Vanderbilt University, 2003) to calculate the effective dose per organ as well as the whole-body effective dose (mSv/MBq) according to ICRP60 and ICRP103 guidelines.
Subsequently, organ segmentations were also performed using Recomia, a cloud-based AI platform for nuclear medicine and radiology research. The workflow for calculating residence times and whole-body effective doses, as described above, was repeated.
Results
Patient data were obtained at three different time-points, day 2, 4, and 7 postinjecton, resulting in 18 PET/CT scans. Overall analysis time was approximately half a workday for manual segmentations compared to ≤30 min using Recomia. Whole-body effective doses differed minimally for the six patients with a median difference in received mSv/MBq of 0.49% (range 0.12 – 1.58%) according to ICRP60 and 0.52% (range 0.15 – 1.95%) according to ICRP103.
Conclusion
These first results suggest that whole-body dosimetry calculations can benefit from fast automated AI based whole-organ segmentations using Recomia. As newly developed MAbs are quickly emerging in anti-cancer therapy, whole-body effective doses for these different therapeutic agents can be assessed quickly and efficiently.