Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

From Monte Carlo PET Simulations to Reconstructed Images : Modelling and Optimisation for 68Ga Theragnostics

Kalaitzidis, Philip LU (2023)
Abstract
In nuclear medicine, radiopharmaceuticals can be administered for both diagnostic and therapeutic purposes. In recent years, there has been an increasing interest in theragnostics, a strategy that combines both diagnosis and therapy. This can be achieved by using similar radiopharmaceuticals for imaging and radionuclide therapy, which enables highly personalised disease management. One theragnostic application is for the diagnosis and management of neuroendocrine tumours, where the diagnosis and subsequent therapy stratification often relies on a qualitative evaluation following [68Ga]Ga-DOTA-TOC PET imaging, with [177Lu]Lu-DOTA-TATE radionuclide therapy being a potential treatment option. In this case,... (More)
In nuclear medicine, radiopharmaceuticals can be administered for both diagnostic and therapeutic purposes. In recent years, there has been an increasing interest in theragnostics, a strategy that combines both diagnosis and therapy. This can be achieved by using similar radiopharmaceuticals for imaging and radionuclide therapy, which enables highly personalised disease management. One theragnostic application is for the diagnosis and management of neuroendocrine tumours, where the diagnosis and subsequent therapy stratification often relies on a qualitative evaluation following [68Ga]Ga-DOTA-TOC PET imaging, with [177Lu]Lu-DOTA-TATE radionuclide therapy being a potential treatment option. In this case, peri-therapeutic SPECT imaging enables for the disease to be closely monitored during therapy. There is growing interest in utilising quantitative metrics to identify the most suitable candidates for radionuclide therapy and to subsequently perform individualised dosimetry. Consequently, it is important to understand potential limitations in the image acquisition process that will impact the accuracy and precision of quantitative estimates, and one effective method to do so is through Monte Carlo simulations.
This thesis is based on four papers utilising Monte Carlo simulations, with a focus on modelling and optimising for 68Ga-PET theragnostics. Paper I explores the possibility of modelling and simulating a clinical GE Discovery MI PET system and coupling simulated data with a reconstruction software, entirely in silico, to enable further simulation-based studies. The implementation of correction factors emulates the processes used in clinical scanners for a more realistic approach. The model successfully generates results comparable to those obtained from a corresponding measurement on a clinical scanner. Papers II, III, and IV focus on 68Ga-PET imaging of neuroendocrine tumours, with Papers III and IV also incorporating 177Lu-SPECT imaging. Anthropomorphic phantoms were utilised to enable the simulation of [68Ga]Ga-DOTA-TOC PET and [177Lu]Lu-DOTA-TATE SPECT exams with patient-like geometries and activity distributions. In Paper II, it was shown that a non-linearly scaled administered activity based on patient weight harmonises image quality, regardless of patient body size. A harmonised image quality is important to ensure that all patients receive an equal standard of care. Paper III investigated the potential impact of respiration on quantitative estimates in [68Ga]Ga-DOTA-TOC PET and [177Lu]Lu-DOTA-TATE SPECT imaging. The extent of lesion motion substantially influenced the recovered lesion activity concentration, with deviations exceeding 30% from the simulated activity concentration. Furthermore, differences in quantitative bias were observed between PET and SPECT imaging, primarily attributed to the different imaging time points. In Paper IV, efforts were undertaken to elevate the realism of simulated patient models, enabling the creation of highly realistic simulated images. The ability to generate realistic images holds great future potential, as it allows for the construction of databases of simulated reconstructed images with known ground truth. These databases can serve various purposes, including software performance evaluation and integration with machine learning. In conclusion, the use of a computational pipeline that connects Monte Carlo simulations with a reconstruction software enables simulation-based studies of entire PET-exam procedures to be conducted. Access to the underlying data driving the simulations makes it possible to isolate individual parameters and track their impact on the results, allowing for a systematic evaluation of in vivo confounders entirely in silico. (Less)
Abstract (Swedish)

Please use this url to cite or link to this publication:
author
supervisor
opponent
  • Professor Lubberink, Mark, Institutionen för kirurgiska vetenskaper, Radiologi, Uppsala universitet.
organization
publishing date
type
Thesis
publication status
published
subject
keywords
Anthropomorphic phantoms, Monte Carlo modelling, Nuclear medicine, PET, SPECT, Theragnostics, Tomographic reconstruction, Fysicumarkivet A:2023 Kalaitzidis
pages
154 pages
publisher
Tryckeriet i E-huset, Lunds universitet
defense location
Föreläsningssal F1, Centralblocket, Skånes Universitetssjukhus Lund, Entregatan 7, 222 42 Lund.
defense date
2023-12-01 09:00:00
ISBN
978-91-8039-878-7
978-91-8039-879-4
language
English
LU publication?
yes
id
d9af6d00-aa54-4ded-89a2-dd3ff4ec9f64
date added to LUP
2023-11-06 12:42:53
date last changed
2024-01-31 10:56:10
@phdthesis{d9af6d00-aa54-4ded-89a2-dd3ff4ec9f64,
  abstract     = {{In nuclear medicine, radiopharmaceuticals can be administered for both diagnostic and therapeutic purposes. In recent years, there has been an increasing interest in theragnostics, a strategy that combines both diagnosis and therapy. This can be achieved by using similar radiopharmaceuticals for imaging and radionuclide therapy, which enables highly personalised disease management. One theragnostic application is for the diagnosis and management of neuroendocrine tumours, where the diagnosis and subsequent therapy stratification often relies on a qualitative evaluation following [<sup>68</sup>Ga]Ga-DOTA-TOC PET imaging, with [<sup>177</sup>Lu]Lu-DOTA-TATE radionuclide therapy being a potential treatment option. In this case, peri-therapeutic SPECT imaging enables for the disease to be closely monitored during therapy. There is growing interest in utilising quantitative metrics to identify the most suitable candidates for radionuclide therapy and to subsequently perform individualised dosimetry. Consequently, it is important to understand potential limitations in the image acquisition process that will impact the accuracy and precision of quantitative estimates, and one effective method to do so is through Monte Carlo simulations.<br/>This thesis is based on four papers utilising Monte Carlo simulations, with a focus on modelling and optimising for <sup>68</sup>Ga-PET theragnostics. Paper I explores the possibility of modelling and simulating a clinical GE Discovery MI PET system and coupling simulated data with a reconstruction software, entirely in silico, to enable further simulation-based studies. The implementation of correction factors emulates the processes used in clinical scanners for a more realistic approach. The model successfully generates results comparable to those obtained from a corresponding measurement on a clinical scanner. Papers II, III, and IV focus on <sup>68</sup>Ga-PET imaging of neuroendocrine tumours, with Papers III and IV also incorporating <sup>177</sup>Lu-SPECT imaging. Anthropomorphic phantoms were utilised to enable the simulation of [<sup>68</sup>Ga]Ga-DOTA-TOC PET and [<sup>177</sup>Lu]Lu-DOTA-TATE SPECT exams with patient-like geometries and activity distributions. In Paper II, it was shown that a non-linearly scaled administered activity based on patient weight harmonises image quality, regardless of patient body size. A harmonised image quality is important to ensure that all patients receive an equal standard of care. Paper III investigated the potential impact of respiration on quantitative estimates in [<sup>68</sup>Ga]Ga-DOTA-TOC PET and [<sup>177</sup>Lu]Lu-DOTA-TATE SPECT imaging. The extent of lesion motion substantially influenced the recovered lesion activity concentration, with deviations exceeding 30% from the simulated activity concentration. Furthermore, differences in quantitative bias were observed between PET and SPECT imaging, primarily attributed to the different imaging time points. In Paper IV, efforts were undertaken to elevate the realism of simulated patient models, enabling the creation of highly realistic simulated images. The ability to generate realistic images holds great future potential, as it allows for the construction of databases of simulated reconstructed images with known ground truth. These databases can serve various purposes, including software performance evaluation and integration with machine learning. In conclusion, the use of a computational pipeline that connects Monte Carlo simulations with a reconstruction software enables simulation-based studies of entire PET-exam procedures to be conducted. Access to the underlying data driving the simulations makes it possible to isolate individual parameters and track their impact on the results, allowing for a systematic evaluation of in vivo confounders entirely in silico.}},
  author       = {{Kalaitzidis, Philip}},
  isbn         = {{978-91-8039-878-7}},
  keywords     = {{Anthropomorphic phantoms; Monte Carlo modelling; Nuclear medicine; PET; SPECT; Theragnostics; Tomographic reconstruction; Fysicumarkivet A:2023 Kalaitzidis}},
  language     = {{eng}},
  publisher    = {{Tryckeriet i E-huset, Lunds universitet}},
  school       = {{Lund University}},
  title        = {{From Monte Carlo PET Simulations to Reconstructed Images : Modelling and Optimisation for <sup>68</sup>Ga Theragnostics}},
  url          = {{https://lup.lub.lu.se/search/files/163947398/Kappa_PhilipKalaitzidis.pdf}},
  year         = {{2023}},
}