Studying Metabolic Brain Connectivity Using 2-Deoxy-2-[18F]Fluoro-D-Glucose Dynamic Positron Emission Tomography at the Single-subject Level

Summary

The acquisition of dynamic positron emission tomography (PET) data and reconstruction into time frames allows for metabolic brain connectivity analysis at the single-subject level. We describe a method to acquire [¹⁸F]FDG dynamic PET data of the rat brain and obtain a connectivity matrix through the extraction of time-activity curves of volumes of interest.

Abstract

To this day, metabolic brain connectivity is mostly studied on a group level through the acquisition of static positron emission tomography (PET) data of multiple subjects. Our research groups are currently studying changes in metabolic connectivity across multiple time points following an intracerebral hemorrhage on an intrasubject level in rats. To investigate intrasubject metabolic brain connectivity, temporal information of the tracer uptake in different brain regions is required, which can be achieved through dynamic PET. In this publication, we give a detailed description of our data acquisition and analysis protocol.

Dynamic PET data of the rat brain are acquired on a dedicated preclinical PET system using 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG) as tracer. The tracer is injected intravenously as a bolus at the start of the PET scan. During the 60 min acquisition, animals are sedated with medetomidine.

After acquisition, the PET data are reconstructed into thirty 2 min time frames using an iterative reconstruction algorithm (Maximum-Likelihood Expectation-Maximization). A parcellated atlas consisting of multiple volumes of interest (VOIs) is used to extract time-activity curves of each VOI, which are then used to calculate the Pearson correlation coefficient between each pair of VOIs.

This dynamic PET protocol enables the assessment of metabolic connectivity differences between two single scans, rather than between groups of scans. This approach allows for the study of changes in metabolic connectivity within a single subject across different time points, or for the comparison of an individual's metabolic connectivity to a normal database. Such comparisons could be useful for tracking disease progression or aiding in the diagnosis of neurological disorders characterized by disrupted communication between brain regions, such as epilepsy or dementia.

Introduction

Positron emission tomography (PET) is a molecular imaging technique commonly used in research as well as in clinical settings. Due to the development of various PET tracers, PET can be used to study disease pathophysiology and monitor disease progression and response to treatments¹. One of the most widely used radiotracers is 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG), which allows for the imaging of glucose metabolism, indicative of cellular activation. It is utilized in oncology for diagnosis, staging, and prognosis; in neurology, commonly in the context of neurodegenerative diseases such as dementia; and in cardiology, to diagnose conditions like sarcoidosis, to name just a few examples^1,2,3.

Assessment of metabolic brain connectivity, obtained from [¹⁸F]FDG PET data, refers to the functional relationships between tracer uptake in different brain regions. This approach enables the computation of a "connectivity matrix" by selecting a set of brain regions, which can provide insights into how different parts of the brain interact and function together. This type of analysis is particularly useful for studying brain function in health and disease, including conditions such as dementia, epilepsy, and other neurological disorders^4,5.

The first study assessing metabolic brain connectivity already dates back to the 1980s⁶, but researchers mainly explored structural brain connectivity, also known as the "connectome"⁷, by means of diffusion-weighted magnetic resonance imaging (DW-MRI). Furthermore, functional connectivity using techniques such as functional MRI (fMRI), electroencephalography (EEG), and magnetoencephalography (MEG) has been widely investigated for multiple decades^8,9.

Recently, there is a regained interest in studying metabolic brain connectivity using [¹⁸F]FDG PET, not only on its own, but also in combination with other forms of brain connectivity¹⁰. However, due to the inherent "static" nature of PET images (in contrast to, e.g., functional MRI), the vast majority of brain network PET-based results are based on group-level analysis, where correlations between brain regions are calculated at the intersubject level. This limitation makes a within-subject analysis of PET images impossible, which is essential for longitudinal studies that can track changes over time within the same individual⁴. Therefore, the development of methods that allow single-subject analysis, such as dynamic PET-based molecular connectivity, is an important research direction in brain research investigating network disorders, since it opens the door to the use of molecular network analysis in clinical practice. Hence, dynamic PET data were used in our preclinical study.

Our research groups are currently conducting a study that examines changes in metabolic connectivity following an intracerebral hemorrhage on an intrasubject level across multiple time points using the rat collagenase model¹¹. To investigate intrasubject metabolic brain connectivity, temporal information of the tracer uptake in different brain regions is required, which can be obtained through dynamic PET. In the following sections, we give a detailed description of the data acquisition and analysis protocol.

Protocol

All procedures are in accordance with the European guidelines (directive 2010/63/EU), and the protocol was approved by the local Animal Ethical Committee of Ghent University (ECD 23/33). Twelve Sprague Dawley rats (six female, six male) were included in the study. Their PET scans were obtained using the following protocol at multiple time points ranging from 2 weeks before to 18 weeks after an induced intracerebral hemorrhage. At the time of the first scan, all animals were 18 weeks of age and the females weighed 244.8 ± 10.1 g (mean ± SD) while males weighed 363.6 ± 13.3 g.

Ensure that radioactive materials are only worked with and handled by trained personnel. Keep the dose to staff and animals as low as reasonably achievable (ALARA).

1. Data acquisition

NOTE: See the Table of Materials for details about the preclinical PET imager and software used for data acquisition. The imager is a 45 detector (arranged in 5-rings) PET scanner covering an axial field of view (FOV) of 13 cm and a transaxial FOV of 7.6 cm, using LYSO crystals and SiPM detectors. The system shows a spatial resolution of 850 µm, a sensitivity of 12%, and an energy resolution of 12.6%¹². The following steps were written with this in mind.

1. Food-deprive the animals for 6-14 h prior to scanning by removing all food from their cage. Once the animal has undergone a sufficient fasting period, start preparations for data acquisition as follows.
2. Turn on a heating pad and allow it to warm up to 30–35 °C.
3. Anesthetize the animal with isoflurane by placing the animal in a gas anesthesia chamber filled with 5% isoflurane in medical O₂ at a flow rate of 2,000 mL/min. Help maintain the animal's body temperature by placing a heating lamp 30 cm from the chamber. Once the respiration rate has decreased to 0.5–1 breaths per second, confirm adequate anesthesia depth by checking for the absence of a paw withdrawal reflex in response to pressure applied to the paw pad.
4. Weigh the animal.
5. Place the animal on the heating pad. Use a nose cone to administer 2% isoflurane in medical O₂ at a flow rate of 500 mL/min. Adjust the isoflurane dose as necessary to keep the animal properly anesthetized. Apply ocular lubricant to prevent corneal drying.
6. Shave a section of the animal's neck between the shoulder blades.
7. Tail vein preparation for tracer injection:
   1. Turn the animal on its left or right side and use the heating lamp to warm its tail, causing vasodilation.
   2. Take approximately 20 cm of BTPE-10 tubing. Using a needle holder, break off the shaft of a 30 G needle and insert 1–2 mm of its back end into the PE tubing. Fill a 1 mL insulin syringe with saline and insert the needle tip 3–5 mm into the PE tubing. Prefill the PE tubing with saline from the filled syringe, and keep the saline syringe connected to the PE tubing.
   3. Clean the animal's tail with ethanol. Remove the heating lamp.
   4. Insert the needle tip of the PE tubing into a lateral tail vein. Confirm proper placement of the needle: look for blood entering the PE tubing when the plunger is pulled back, an absence of resistance while injecting, and the formation of subcutaneous bleb (indicating paravenous injection) at the injection site. If the entry has not been successful, remove the needle from the tail and try again at a more proximal section.
   5. Secure the needle with a drop of instant glue at the penetration site and tape the PE tubing to the tail.
      ​NOTE: Avoid movement of the PE tubing. If at any moment blood enters the PE tubing, flush it immediately with saline to prevent clogging.
8. Switch to medetomidine sedation:
   1. Place the animal in a prone position on the heating mat. Reapply ocular lubricant if necessary.
   2. Prepare PE tubing (40–100 cm) to connect to an infusion pump that is placed close to the PET scanner. Using a needle holder, break off the shaft of a 30 G needle and insert 1–2 mm of its back into the PE tubing. Fill a 1 mL plastic syringe with a 30 G needle with at least 0.2 mL medetomidine (per animal), and insert the needle tip 3–5 mm into the PE tubing. Prefill the PE tubing with medetomidine from the syringe.
   3. Place the syringe with medetomidine in the infusion pump. Turn on the pump and select the correct syringe by clicking Find syringe | BD Plastic | 1 mL.
   4. Eject boli of 0.005 mL using the pump until a drop emerges from the needle at the other end of the PE tubing. Ensure no air bubbles are present in the PE tubing.
   5. Insert the needle of the PE tubing 2–3 mm subcutaneously between the shoulder blades of the animal. Secure the needle with a drop of instant glue at the penetration site and gently tape the PE tubing to the animal's back.
   6. Inject a bolus of volume v of medetomidine (rate: 3 mL/min) based on the animal's weight:
      
      Where d = 0.05 mg/kg (the dose of medetomidine in mg per kg of animal weight), m is the animal's weight in kg, and c is the concentration of medetomidine in mg/mL.
   7. Stop isoflurane administration 10 min after the administration of the medetomidine bolus, adjusting the isoflurane dose as necessary to maintain stable anesthesia during this waiting period.
      NOTE: Isoflurane combined with medetomidine significantly slows the animal's respiratory rate. Close monitoring is essential until isoflurane administration ceases.
   8. Five minutes after stopping isoflurane, begin a continuous infusion of medetomidine at a rate of:
      r_cont = r • m
      Where r = 0.1 mg/(kg·h); the amount of medetomidine in mg per animal body weight in kg/h), and m is the animal's weight in kg.
      ​NOTE: An equilibration period of an additional 25 min after starting continuous infusion (thus 30 min after stopping isoflurane) is needed before the scan can be initiated to ensure the animal's system is free of isoflurane. During this time, proceed with the following steps.
9. Animal positioning in the PET scanner:
   1. Prepare the scanner bed with a heating pad, a pressure sensor, and paper tissues. Use the sensor to monitor the animal's respiratory rate and the paper tissues for absorbing urine excreted by the animal.
      NOTE: Due to the medetomidine, the animal will typically excrete large amounts of urine.
   2. Carefully transfer the animal onto the scanner bed, ensuring the medetomidine line and tail vein catheter remain secure. Position the animal's head as straight as possible using a nose cone for assistance if necessary. Secure the head with tape once properly positioned.
   3. Turn on the heating pad and verify the respiratory rate monitoring system is functioning correctly, adjusting pressure sensor positioning if needed.
10. Ten to five minutes before starting the PET scan, fill an insulin syringe with approximately 20 MBq 2-deoxy-2-[¹⁸F]fluoro-D-glucose ([¹⁸F]FDG) diluted in 0.1–0.8 mL of saline.
    ​CAUTION: Fluorine-18 (¹⁸F) is a fluorine radioisotope with a half-life of 109.8 min. Follow local radiation safety protocols when handling radioactive material.
11. Open the scanner software and configure the following scan parameters: Select ¹⁸F as the isotope (important for decay correction), enter the measured activity of the syringe containing the radiotracer (in MBq measured by the dose calibrator), and the time of measurement (the information will be saved in the metadata, which is needed to convert the data to other units such as standard uptake values). Select the General protocol and tick Automatic sorting after acquisition. Set the scan duration to 1 h, and adjust the region of interest (ROI) so it contains the entire rat brain (typically position 2-5). 
12. Once 30 min have passed since the stop of isoflurane administration, initiate the PET scan:
    1. Again, confirm proper placement in the vein by aspiration followed by injection of a small amount of saline as described in step 1.7.4.
    2. Replace the saline syringe with the radiotracer syringe. When removing the syringe from the PE tubing, it is possible that blood enters the PE tubing, pushing out saline. Insert the needle of the radiotracer syringe quickly without puncturing the PE tubing.
    3. Start the scan using the scanner software.
    4. Once the acquisition begins, inject the tracer as a bolus over approximately 0.5–2 s. Immediately replace the empty radiotracer syringe with the saline syringe and flush the PE tubing with approximately 0.1 mL of saline. Use a tissue to catch any drops of radiotracer that may escape during the syringe swap.
       NOTE: The count rate displayed by the scanner software should rapidly increase to 4.5–6.5 Mcps following tracer injection. The count rate initially being significantly below 4 Mcps and increasing afterwards indicates a (partially) paravenous tracer injection.
    5. Assess the remaining activity that was not injected into the animal by measuring the activity of the tracer syringe and potentially contaminated tissue(s) and gloves. Enter that activity and the time of measurement into the scanner software to have it saved in the metadata.
       ​NOTE: A remaining activity of 0.1–1.5 MBq is typical, depending on how much tracer escaped when switching the tracer for the saline syringe.
13. Once the scan is completed, eject the bed with the animal from the scanner.
    NOTE: The animal may react to stimuli or wake up before medetomidine has been antagonized. Handle the animal with care and be prepared for sudden movements during the following steps. Also be aware that the animal is radioactive. Thus, aim to complete the following steps as quickly and carefully as possible.
14. Remove the needle and PE tubing from the animal's tail using a needle holder. Apply a tissue or compress to the puncture site to stop potential bleeding.
15. Stop medetomidine administration and remove the needle and PE tubing from the animal's neck.
16. Place the animal in a cage free of other animals, using a tissue to catch (radioactive) urine that may be excreted while moving the animal. Administer 1 mL of saline subcutaneously in its back using a 1 mL insulin syringe to promote rehydration.
17. Reverse the sedation with a subcutaneous injection of atipamezole.
    1. Calculate the appropriate volume v using the following equation:
       
       Where d = 0.5 mg/kg, i.e. the dose of medetomidine in mg per kg of animal weight, m is the weight of the animal in kg, and c is the concentration of atipamezole in mg/mL.
    2. For light-weight animals resulting in small calculated volumes, dilute the atipamezole with saline to allow for a more accurate dose. Fill an insulin syringe with 4 • v of atipamezole and 12 • v of saline. Inject 1/4 of the resulting solution subcutaneously into the animal's back near the neck.
18. Monitor the animal visually until it wakes up, which should take 1–5 min. Once the animal is able to keep itself upright and move around freely, return it to group housing. Provide the animal with food and water.

2. Data reconstruction and quality check

NOTE: Using the hard- and software available in our study, all PET data were corrected for radionuclide decay, and acquired sinograms were reconstructed with the ordered subset expectation maximization 3-dimensional (OSEM-3D) algorithm using a 30% window around the 511 keV photopeak. The OSEM reconstruction software used its default settings of 10 subsets with 20 million events per subset. Images were reconstructed into a 192 x 192 x 384 transverse matrix with cubic image voxels of 0.4 mm. No attenuation correction was performed.

1. To verify that tracer injection was successful, check the count rate detected during the 60 min acquisition (i.e., time-activity curve (TAC)). Confirm that there is a peak within the first minute of acquisition, immediately after the tracer had been injected.
2. Static reconstruction: Reconstruct the acquired data using an OSEM algorithm using 30 iterations, an isometric voxel size of 400 µm, and an energy window of 30% centered on the 511 keV photopeak.
3. Once the reconstruction is complete, inspect the resulting image for artifacts.
4. Dynamic reconstruction: Reconstruct the acquired data into thirty 2 min time frames using the same settings as for the static reconstruction.

3. Data analysis

NOTE: The following steps 3.1 and 3.2 are performed in a biomedical software environment dedicated to quantification of PET data, using the Image Registration and Fusion Tool (PFUS) and the General Kinetic Modeling Tool (PKIN).

1. In PFUS:
   1. In sub-page Load Input Images:
      1. Load the DICOM file of the dynamic reconstruction by either using the INP load button or by dragging and dropping the file into the canvas. Use the General image display control in the right top to navigate through slices and time frames and adjust the color bar. If needed, adjust the image orientation to standard orientation by using the Change input image orientation button on the right sidebar.
      2. Cropping and frame average: Click the arrow in the right bottom to display the hidden controls for summing and cropping and tick Crop. In the canvas, define the boundaries for cropping by adjusting the edges and dragging the center of the cropping box; ensure the whole brain and a bit of padding, containing parts of the eyes and spinal cord, are contained. Once the cropping box is defined properly, click the cropping button to crop the image. Calculate the time-weighted frame average over all time frames by clicking the orange average button and wait for the result to be automatically added to the input images, resulting in two input images in total: the cropped dynamic data with the time frames and the cropped time-weighted average of the same data.
      3. Select Deform as Matching method or press the button if already selected to move on to template-based normalization.
   2. In sub-page Reference & Matching:
      1. Select the time-weighted average image as the input image (it is normally selected automatically from the last step).
      2. Load a reference by clicking on the button to select a normalization template as the reference image. Select Rat FDG (W. Schiffer).
      3. Match the input to the template by clicking the button for Input matching | Input: Adjust manually in the sidebar on the right. Drag and rotate the input image to roughly match the template. Ensure that Rat is selected as species, the registration method is Deform, then click the Match Current button to reach the next sub-page.
   3. In sub-page Matching Result:
      1. Check the result and adjust the match manually if needed. Click Apply Current Transformation to All to also match the input image consisting of the time frames.
      2. Click the green button VOIs to move on to the next processing step.
   4. In sub-page VOI Analysis:
      1. Select the image containing the time frames as image A in the bottom right menu.
      2. Go to the Template tab and select Px Rat (W. Schiffer) in the Atlas tab, or load a custom atlas, and select the desired volumes. To apply the VOIs, click the Outline button in the bottom.
      3. Click the button above the Template tab to calculate TACs and send them to PKIN tools.
2. In PKIN:
   1. In the bottom right, select TACs for Display Type to visualize the TACs in kBq/mL of the VOIs selected beforehand.
   2. Right-click on the canvas with the TACs and click Value table of visible curves. Copy the table to a spreadsheet for further analysis.
3. Calculate the Pearson correlation between all pairings of VOIs, using data from the sixth to last time frame (= 10–60 min post scan start), to obtain a correlation matrix. A Python script called “correlation_matrix.py” fulfilling this purpose is included in Supplemental File 1.
4. Calculate whether the Pearson correlation is significant by using the t-test statistic:  with n - 2 degrees of freedom, where r is the Pearson correlation coefficient and n = 25 the number of time frames used for calculating the Pearson correlation. Apply a suitable correction for multiple comparisons and compare the statistic to the critical value of a two-tailed t-distribution (p = 0.05) with n - 2 degrees of freedom. Consider the Pearson correlation between two VOIs significant if the corrected t-test value is smaller than the critical value.

Results

Once the scan is completed, the TAC of the detected rate during the acquisition can be investigated to check for a correct tracer injection and uptake. Figure 1 displays a TAC resulting from the whole FOV of the scanner after a successful tracer injection and acquisition (panel A), and a TAC resulting after a partially paravenous tracer injection (panel B). In the successful case, the count rate rises rapidly after tracer injection and reaches its peak within the first 4 min...

Discussion

The protocol provided here guides users through the process of acquiring 1 h dynamic PET data using [¹⁸F]FDG as a tracer in rats. In the end, a correlation matrix of VOIs is obtained, which can be used to assess metabolic connectivity on a single-subject level. Experienced researchers may adjust the protocol to fit their specific needs at various points, for example, by using a different radiotracer, acquisition time, or time frame widths for image reconstructions, and selecting relevant VOIs in the data analy...

Materials

List of materials used in this article

Name	Company	Catalog Number	Comments
Antisedan	Orion Pharma		Atipamezole hydrochloride 5 mg/mL
BD Micro-Fine+ insulin syringe 1 mL	BD	324827	0.33 mm (29G) x 12.7 mm
BD Microlance 3 Needles 30 G x 1/2"	BD	304000	30 G x 1/2"; 0,3 x 13 mm
BD Plastipak syringe 1 mL	BD	303172	for infusion pump
BTPE-10 Polyethylene tubing	Instech		0.11x.024in (.28x60mm)
Domitor	Orion Pharma	1070499	Medetomidine hydrochloride 1 mg/mL
Fusion 100 infusion pump	Chemyx Inc.	07100	Newer model available: Fusion 100X
Isoflutek 1000 mg/g	Alivira		Isoflurane
MOLECUBES β-CUBE with CUBEFLOW software	MOLECUBES NV		Preclinical PET scanner
PMOD Software version 4.4	Bruker Corporation		http://www.pmod.com; quantification of PET data
Saline	B. Braun	394496	NaCl 0.9%
Vidisic eye gel	Vidisic		Carbomerum 980 2 mg/g