Periodic Reporting for period 1 - simpetmr (Development of quantification procedures for simultaneous PET-MR data for human brain imaging)
Reporting period: 2019-03-11 to 2021-03-10
The discovery and development of drugs for treatment of brain disorders is an enormously challenging process requiring large resources, timelines, and associated costs. Neuroimaging with Positron Emission Tomography (PET) has become a central component of the evaluation of novel drugs for brain disorders by facilitating decision-making already in phase I studies, and thereby provides the possibility to discharge risks very early in the developmental pipeline. With PET, researchers can demonstrate drug penetration and kinetics in the brain and quantify pharmacodynamic effects. This is typically based on experiments where the occupancy of the drug is determined, i.e. the fraction of available binding sites occupied by the drug at a certain dose.
Drug occupancy is traditionally estimated by performing multiple PET measurements before and after administration of a drug. Neuroimaging with PET is a remarkably expensive research tool, which often puts a significant strain on the research budget. It also involves exposing research subjects to ionizing radiation, which motivates why we want to minimize the number of PET scans acquired, without sacrificing the knowledge they provide. In this research project, I have developed mathematical models that allows researchers to estimate drug occupancy from a single scan, thereby enabling that the number of PET scans taken are reduced by half.
The idea is to administer the drug during an on-going scan (referred to as a displacement scan). As the drug enters the brain, it will displace the tracer molecule, causing a perturbation in the measured signal. With the models I have developed, this perturbation is analysed in such a way that useful information regarding the drug’s binding in brain can be gained.
A secondary objective of this research was to develop similar models for simultaneous PET/MR acquisitions. Whereas PET provides information on brain neurochemistry, functional magnetic resonance imaging (fMRI) can provide information on changes in cerebral blood flow that results as a downstream effect from the drug-receptor interaction. Thus, performing displacement scans in a simultaneous PET/MR setting would allow researchers to obtain both these pieces of information at the same time.
The overall conclusions from this work is that the mathematical models I have developed constitute a valuable tool for researchers interested in studying the interactions between a drug and brain tissues. They are a set of easy-to-use functions that are anchored in our current understanding of radioligand kinetics in brain tissue. The validation work performed in this project confirm that unbiased and accurate estimates of drug occupancy can be obtained from a single PET scan, given that a sufficiently large dose is given. For smaller doses, although the estimates are unbiased, they come with some uncertainty. Unfortunately, I was not able finalize the development of the models for simultaneous PET/MR acquisitions. At the point in time where this development was about to start, the pandemic was in full bloom. As the health care sector was suffering, access to scanner slots and radioligand production for all non-clinical activities was heavily restricted, and as an effect, I could not perform any PET/MR acquisitions.
Drug occupancy is traditionally estimated by performing multiple PET measurements before and after administration of a drug. Neuroimaging with PET is a remarkably expensive research tool, which often puts a significant strain on the research budget. It also involves exposing research subjects to ionizing radiation, which motivates why we want to minimize the number of PET scans acquired, without sacrificing the knowledge they provide. In this research project, I have developed mathematical models that allows researchers to estimate drug occupancy from a single scan, thereby enabling that the number of PET scans taken are reduced by half.
The idea is to administer the drug during an on-going scan (referred to as a displacement scan). As the drug enters the brain, it will displace the tracer molecule, causing a perturbation in the measured signal. With the models I have developed, this perturbation is analysed in such a way that useful information regarding the drug’s binding in brain can be gained.
A secondary objective of this research was to develop similar models for simultaneous PET/MR acquisitions. Whereas PET provides information on brain neurochemistry, functional magnetic resonance imaging (fMRI) can provide information on changes in cerebral blood flow that results as a downstream effect from the drug-receptor interaction. Thus, performing displacement scans in a simultaneous PET/MR setting would allow researchers to obtain both these pieces of information at the same time.
The overall conclusions from this work is that the mathematical models I have developed constitute a valuable tool for researchers interested in studying the interactions between a drug and brain tissues. They are a set of easy-to-use functions that are anchored in our current understanding of radioligand kinetics in brain tissue. The validation work performed in this project confirm that unbiased and accurate estimates of drug occupancy can be obtained from a single PET scan, given that a sufficiently large dose is given. For smaller doses, although the estimates are unbiased, they come with some uncertainty. Unfortunately, I was not able finalize the development of the models for simultaneous PET/MR acquisitions. At the point in time where this development was about to start, the pandemic was in full bloom. As the health care sector was suffering, access to scanner slots and radioligand production for all non-clinical activities was heavily restricted, and as an effect, I could not perform any PET/MR acquisitions.
This project incorporated both theoretical and practical parts. Initially, mathematical models were derived that allowed for quantification of PET data which include a perturbation. This represent the development of a whole new class of models, because all existing models for PET quantification relies on the assumption of a “steady-state” (more technically, they model a time-invariant system). I therefore revisited the most commonly used PET kinetic models, and reformulated them so that they no longer relied on a steady-state. Extensive simulation experiments were conducted, and the models were continuously improved and tuned, so that their predictive power was maximized.
The second step involved the acquisition of suitable data. Here, I had to deviate from the plan set out in the proposal, as I gradually understood that my original plan would not work. More specifically, my initial experiments revealed the tracer I intended to use could not be displaced by subsequent injection of a drug. Up until that point, such behaviour had previously been reported in neuroscience literature, and it itself sparked a new line of research at the lab. Yet, for the purpose of my displacement models, this meant that I had to change from studying the serotonergic neurotransmission system, and instead I focused on a newly developed PET tracer that targets a protein that is abundant in the synapses. My initial experiments showed that this tracer could be very well displaced by epilepsy medicine (Brivaracetam and Levetiracetam). This had two unforeseen consequences: 1) It was very late during my fellowship that we received approval to use that tracer in humans, and as such I was forced to conduct all experiments in pigs. 2) This new experimental setup is very valuable for epilepsy research, as my models became a tool that can be used to examine why some epilepsy patients don’t respond well to their medication.
The results achieved so far thus consists of a set of models that can be used to quantify drug interaction effects on the living human brain from a single PET measurement. The models are thoroughly validated through the use of data simulations and PET displacement scans acquired in pigs. To perform the equivalent work in a simultaneous PET/MR setting remains to be done.
The second step involved the acquisition of suitable data. Here, I had to deviate from the plan set out in the proposal, as I gradually understood that my original plan would not work. More specifically, my initial experiments revealed the tracer I intended to use could not be displaced by subsequent injection of a drug. Up until that point, such behaviour had previously been reported in neuroscience literature, and it itself sparked a new line of research at the lab. Yet, for the purpose of my displacement models, this meant that I had to change from studying the serotonergic neurotransmission system, and instead I focused on a newly developed PET tracer that targets a protein that is abundant in the synapses. My initial experiments showed that this tracer could be very well displaced by epilepsy medicine (Brivaracetam and Levetiracetam). This had two unforeseen consequences: 1) It was very late during my fellowship that we received approval to use that tracer in humans, and as such I was forced to conduct all experiments in pigs. 2) This new experimental setup is very valuable for epilepsy research, as my models became a tool that can be used to examine why some epilepsy patients don’t respond well to their medication.
The results achieved so far thus consists of a set of models that can be used to quantify drug interaction effects on the living human brain from a single PET measurement. The models are thoroughly validated through the use of data simulations and PET displacement scans acquired in pigs. To perform the equivalent work in a simultaneous PET/MR setting remains to be done.
During the past decade, imaging with PET has positioned itself as an invaluable tool to determine properties of new drugs, and is frequently used for drug development purposes. A limiting factor in close to all research is the amount of ionizing radiation and the costs associated with these experiments. In addition, performing multiple PET experiments on the same individual comes with logistic difficulties, and in many cases the results may be influenced by changes in diet, sleep or exercise between the imaging session, which may cause some data to be difficult to interpret. My models thus present a novel tool to simplify these types of examinations. I have already been in contact by imaging CROs whose shown interest in using this experimental setup for their future studies. Combining these models to a simultaneous PET/MRI paradigm is a promising avenue, and ideally researchers would use MR imaging alone to study drug occupancy, and thereby circumvent the challenges associated with PET. In such context, my models and the experimental setup proposed in this project will serve as a gold standard, to which simplified experimental procedures can be compared.