Accuracy of First Day Few-Time-Points 131I Dosimetry for Thyroid Disease using Nonlinear Mixed-Effects Modelling
Moh Faizal Fajri Al Amin (a-),Teuku M. Zaki Yasykur Polem (b-), Indra Budiansah(b), Idham Arief (a), Freddy Haryanto(a), Peter Knoll(c#), Deni Hardiansyah(b*), and Gerhard Glatting(d#)

a) Nuclear Physics and Biophysics, Physics Department, Faculty of Mathematics and Natural Sciences, Institut Teknologi Bandung, Bandung, Indonesia, 40132.
b) Medical Physics and Biophysics, Physics Department, Faculty of Mathematics and Natural Sciences, Universitas Indonesia, Depok, Indonesia, 16424.
c) International Atomic Energy Agency (IAEA)
d) Medical Radiation Physics, Department of Nuclear Medicine, Ulm University, Ulm, Germany.

- shared first authorship
# shared senior authorship
* corresponding author

Abstract

Background: Accurate Time-Integrated Activity (TIA) estimation is essential for patient-specific 131I dosimetry, but conventional multi-time-point measurements are clinically demanding. Early few-time-point approaches can reduce this imaging burden. This study evaluated the accuracy of early few-time-point TIA estimation using Non-Linear Mixed-Effects Modelling (NLMEM). Material and Methods: Time-activity data for 800 virtual patients (VPs- 600 training, 200 testing) at 1, 2, 4, 6, and 10 h post-administration were generated utilizing published biokinetic parameters (PMID: 40198532). NLMEM was applied to describe population and individual biokinetics using a published sum-of-exponentials model for 131I therapy. Reference TIAs (rTIAs) for the testing virtual patients (VPs) were obtained from NLMEM fitting using the full available time-point data, with testing and training VPs fitted simultaneously. To evaluate reduced sampling, estimated TIAs (eTIAs) for testing VPs were calculated using subsets of two to five early time points, maintaining simultaneous fitting. Accuracy against rTIAs was assessed via relative deviation (RD), mean absolute percentage error (MAPE), and root-mean-square error (RMSE). Results: Optimal TIA estimation accuracy consistently required the 10-h measurement. Sampling schedules incorporating this time point-ranging from the full early set to a reduced three-point combination (1, 2, and 10 h)-maintained robust predictive performance, yielding RMSE of 51.16%-51.35%, MAPE of 29.96%-30.48%, and mean RD of 18.12%-18.52%. Conversely, excluding the 10-h measurement significantly degraded accuracy, with RMSE and MAPE exceeding 60% and 34%, respectively. Conclusion: The 10-hour post-administration measurement is a critical component for accurate early few-time-point dosimetry in 131I therapy. Combining NLMEM with reduced sampling protocols that include the 10-hour time point can provide reliable TIA estimates while reducing clinical imaging burden.

Keywords: Thyroid Dosimetry, Early Few-Time-Points, Non-Linear Mixed-Effects Modelling (NLMEM)

Topic: Medical Physics and Biophysics