A General-Purpose Adaptable Foundation Model for Total-Body PET Kinetic Modelling
Abstract
Voxel-level kinetic modeling in total-body dynamic positron emission tomography (TB-PET) enables precise physiological quantification but complicates parameter uncertainty estimation. While generative consistency models (CMs) can rapidly estimate Bayesian posterior distributions, they are constrained by type of tracers, kinetic models, and protocols. Training them from scratch whenever a new tracer, model or protocol is used can be computationally intensive. To address this, we introduce LoRACM, a parameter-efficient framework utilizing low-rank adaptation (LoRA) to transfer a pretrained foundation CM to novel TB-PET scenarios. We evaluated three adaptation scenarios using simulated and human data: scanner and protocol transfer for FDG two-tissue compartment modeling; FP-CIT simplified reference tissue modeling; and high-temporal-resolution FDG blood flow imaging via the adiabatic approximation to tissue homogeneity model. Each adaptation was trained in less than 3 hours on a single NVIDIA A100 GPU using 50,000 simulated pairs. In simulations, LoRACM showed strong agreement with results from Markov Chain Monte Carlo (MCMC, which we use as reference posterior distributions), yielding median Kullback-Leibler divergences of 0.05-0.11 and median absolute percentage errors below 1.2% for each kinetic parameter of interest. In human data, LoRACM generated total-body K_i maps comparable to the ones made by non-negative nonlinear least squares (NNLS) fitting and a consistency model trained from scratch. Striatal dopamine binding potential estimates predicted by LoRACM correlated with NNLS estimates (r = 0.968-0.988). LoRACM produced total-body blood flow imaging with FDG matching the one made by NNLS. These underscore the potential of adapting a pretrained CM with LoRA for fast and reliable posterior parameter inference for kinetic models across distinct TB-PET applications.
Related
- GMAP: A General Purpose Hamiltonian Mapping Program for Computational Spectroscopy
- vibe‐qc: An Open‐Source, General‐Purpose Electronic‐Structure Code for Molecules and Solids Developed With Generative AI
- Assessing the Transferability of General-Purpose MachineLearning Interatomic Potentials for Heterogeneous Catalysis with HetCat26
- General-purpose equations of state for hadron-quark matter: first application to accretion-induced phase transitions in neutron stars
- QMClaw: A Scalable General-purpose Framework for Quantum Measurement and Control
BibTeX
@misc{zhao2026generalpurpose,
title = {{A General-Purpose Adaptable Foundation Model for Total-Body PET Kinetic Modelling}},
author = {Yun Zhao and Qinlin Gu and Andrew J. Reader and Yanan Fan and Georgios I. Angelis and Katya Amadita and Kevin J. Chung and Xiaoyu Duan and Benjamin A. Spencer and Guobao Wang and Steven R. Meikle},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.41093},
eprint = {2610.41093},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.41093},
}
RIS
TY - UNPB
TI - A General-Purpose Adaptable Foundation Model for Total-Body PET Kinetic Modelling
AU - Yun Zhao
AU - Qinlin Gu
AU - Andrew J. Reader
AU - Yanan Fan
AU - Georgios I. Angelis
AU - Katya Amadita
AU - Kevin J. Chung
AU - Xiaoyu Duan
AU - Benjamin A. Spencer
AU - Guobao Wang
AU - Steven R. Meikle
PY - 2026
DA - 2026/10/30
PB - arXiv
DO - 10.48550/arxiv.2610.41093
UR - https://arxiv.org/abs/2610.41093
AB - Voxel-level kinetic modeling in total-body dynamic positron emission tomography (TB-PET) enables precise physiological quantification but complicates parameter uncertainty estimation. While generative consistency models (CMs) can rapidly estimate Bayesian posterior distributions, they are constrained by type of tracers, kinetic models, and protocols. Training them from scratch whenever a new tracer, model or protocol is used can be computationally intensive. To address this, we introduce LoRACM, a parameter-efficient framework utilizing low-rank adaptation (LoRA) to transfer a pretrained foundation CM to novel TB-PET scenarios. We evaluated three adaptation scenarios using simulated and human data: scanner and protocol transfer for FDG two-tissue compartment modeling; FP-CIT simplified reference tissue modeling; and high-temporal-resolution FDG blood flow imaging via the adiabatic approximation to tissue homogeneity model. Each adaptation was trained in less than 3 hours on a single NVIDIA A100 GPU using 50,000 simulated pairs. In simulations, LoRACM showed strong agreement with results from Markov Chain Monte Carlo (MCMC, which we use as reference posterior distributions), yielding median Kullback-Leibler divergences of 0.05-0.11 and median absolute percentage errors below 1.2% for each kinetic parameter of interest. In human data, LoRACM generated total-body K_i maps comparable to the ones made by non-negative nonlinear least squares (NNLS) fitting and a consistency model trained from scratch. Striatal dopamine binding potential estimates predicted by LoRACM correlated with NNLS estimates (r = 0.968-0.988). LoRACM produced total-body blood flow imaging with FDG matching the one made by NNLS. These underscore the potential of adapting a pretrained CM with LoRA for fast and reliable posterior parameter inference for kinetic models across distinct TB-PET applications.
SN - 2331-8422
ER -