The first application of positron annihilation medical imaging was first explored in by Brownell et al. for imaging brain tumors.47 This technique was later expanded to tomographic imaging in 1971.48 Myocardial perfusion PET with 13N ammonia was one of the first practical implementations of PET in the myocardium.49 13N ammonia produces high-quality myocardial perfusion images with favorable responses to changes in blood flow.49–50 The short half-life of 13N requires a nearby cyclotron for production, thereby limiting the number of sites that can use it. A more practical approach employs a generator to produce the radionuclide. 82Rb was studied in 1986 as a generator-produced potassium analog for the detection of myocardial infarction.51–52 Using an 82Sr generator (T1/2 = 28 days), a user could be provided with an on-site generator every 4 weeks. More recently, these generators can be supplied for up to 6 weeks, depending upon available camera technology.53,54
Metabolic cardiac imaging with PET has been of great interest because of the relative simplicity of using PET radionuclides with metabolically active molecules. 18F-labeled deoxyglucose (FDG) (a glucose analog) has long been used for imaging myocardial viability55,56 and has been also used for imaging of ischemic activation57 and sarcoid imaging.58 As a metabolic agent, FDG is trapped in cells that utilizes predominately glucose metabolism. This can be present in hibernating myocardium, severely ischemic myocardium, infections and inflammation as well as in tumors. In addition, metabolic agents have been explored for fatty acid metabolism, cardiac denervation, and the detection of amyloid plaques.59–61
PET photon events consist of the following62:
True pairs: Actual coincidence events where two unscattered 511-MeV photons produced by the same positron annihilation and are detected by the system. These events are used for producing the tomographic images.
Photon scatter events: Coincidence events where two photons produced by the same positron annihilation are recorded by the system, but at least one of the photons has been scattered by electrons in the patient. These events degrade image contrast.
Random events: Coincidence events where two photons produced by different positron annihilations are recorded by the system within the timing gate of the system. These events degrade image contrast.
Prompt gamma: Additional photons that are produced during the nuclear decay that produces the positron. Given the timing, these photons can appear as coincidence photons. With Rb-82, 13% of all decays produce and additional 776-keV prompt gamma. When these photons are not accounted for, the over-correction for scatter can reduce specificity from 90% to 22%.63
A coincidence "event" occurs when two 511-keV photons resulting from positron decay and annihilation, emitted nearly 180 degrees apart, are detected "simultaneously" within a small (4–12 nanoseconds) timing window.62 Most PET scanners consist of a system of concentric rings of multiple-detector blocks around a central axis (see Fig. 4-10). Each of these detectors contains a scintillator crystal and a small array of photo-multiplier tubes. The physical properties of the scintillator crystal determine much of the performance of the PET scanner. Ideally, a scintillator crystal has a high stopping power (to reduce the depth the gamma rays penetrate before being recorded), high light output per event (needed to differentiate photon energies), and a rapid light curve decay (to minimize the time needed to record the photon event and have the crystal ready to receive the next event). Response time of the system describes how effective a system can differentiate random events from true coincidence. To accomplish this, a coincidence timing window of <10 nanoseconds is needed to differentiate true pairs from random coincidence event. As challenging as that may appear, most modern PET scanners have coincidence windows of <5 nanoseconds, making it possible to not only differentiate the difference between randoms and trues, but utilize the differences in the arrival times of the two annihilation photons to localize the annihilation event along the line of site.
A PET scanner consists of a concentric ring of detector design to detect the coincidence of photon received by the scanner. A timing gate is used to identify photon events that fall in the correct timing range, approximately 5 to 10 nanoseconds. The logic also determines if the events are within an acceptable energy range.
The intrinsic sensitivity of PET is typically much higher than SPECT due to the fact that localization of the annihilation event does not require the use of collimation. Despite this fact, a large number of PET imaging systems use a system of septa to reduce scatter and randoms and thereby improve the image quality of the final reconstructed images.
Recent enhancements in processing techniques have enabled fully 3D imaging (septa removed) with Rb-82.64 Removing the septa significantly improves the sensitivity of the imaging system, potentially increasing system sensitivity by a factor of two to five times.65 The American Society of Nuclear Cardiology recommends reducing the infused dose of Rb-82 from 50 mCi using two-dimensional (2D) imaging to 30 mCi in 3D. Similar dose reduction is also recommended for FDG imaging.66
Attenuation Correction of PET
In contrast with SPECT, AC is almost always applied in the processing of cardiac PET. The geometry of PET makes it possible to estimate attenuation without knowing the depth of the annihilation event in the patient. Unlike SPECT, the attenuation of a true event is the sum of the attenuation both 511-keV photons; specifically the total attenuation along the line of response (see Fig. 4-11). Therefore, the correction that is applied is independent of where the annihilation event took place along a line of response. The result is a high-resolution final image with greatly reduced attenuation and scatter artifacts (see Fig. 4-12).
Attenuation in PET is different than SPECT because the attenuation experienced by a true pair is the sum of the attenuation of both photons, for example, the total line of response. Because of this, it is not necessary to know the depth of the annihilation in the patient to correct the attenuation.
A high-quality PET perfusion study has a nearly uniform uptake appearance.
Radionuclide Source for Attenuation Correction
Radionuclide sources for transmission imaging used for AC are Germanium-68 (Ge-68, 511 keV) or Cesium-167 (Cs-137, 630 keV). Both are contained in rotating rod sources that can be extended and retracted for image acquisition as needed throughout the study. Such cameras that utilize radioactive line sources are called "dedicated" PET systems. Because the source photons are positrons, there is no need to translate the attenuation coefficients from the transmission energy to the emission energy. Another advantage of dedicated PET is the attenuation map is acquired while the patient is free breathing. This reduces misregistration and breathing artifacts when compared to PET/CT.67,68 Finally, the radiation dosage from a dedicated PET scanner is typically less than an x-ray–based CT (though dosage-reducing strategies can be employed to minimize x-ray radiation dosage). One possible limitation of dedicated PET is some systems require long transmission acquisition scans (4–8 minutes). Newer reconstruction methods with newer instrumentation reduce the number of counts required for image reconstruction, reducing the acquisition time to 60 to 90 seconds.40
CT Source for Attenuation Correction
PET-CT systems can employ a conventional multi-slice CT system, capable of acquiring high-resolution, diagnostic quality data and/or low dose, AC specific x-ray tubes. CT-derived attenuation maps have the advantage of being able to acquire a high signal-to-noise transmission image in a short period of time. Despite this, CT-specific artifacts can make PET/CT more challenging than line source AC (see Fig. 4-13).67 Patient motion and breathing introduce the greatest challenges. Several approaches have been proposed such as end expiration breath holding68 and shallow free breathing. Cine/CT has been demonstrated to be both robust and easy to implement.69 By using several passes over the diaphragm, an average position for the diaphragm can be obtained. Though this does reduce breathing artifacts, it requires rescanning the same tissue areas several times to cover the entire respiratory cycle, thus increasing potential radiation dose. Lowering kVp and tube current can help reduce the dose, however these improvements still result in CTAC scans of over 2 mSv.
Attenuation correction of PET using CT can be complicated in the presence of metal artifacts. Though this is not a contraindication for cardiac PET, corrections must be applied to minimize the influence of metal beam hardening artifacts.
Radiation exposure for the CT portion of a PET/CT scan can be a significant contributor to the overall radiation of the patient study and can be as high as 8 mSv when dose optimization is not used.70 Optimizing the acquisition setting is crucial for obtaining the best transmission study at the lowest possible radiation exposure71:
Optimize the CT scanner acquisition parameters. The x-ray tube voltage (kVp, which determines the mean energy of the x-rays produced by the x-ray tube) and tube current (mA which is proportional to the photon output of the x-ray tube). These should be less than 100 kVP and 10 mAs for most applications
When possible, prospective ECG triggering should be used.
If helical scanning is required, dose modulation should be employed.
Field-of-view settings for obtaining the AC scan should be confined to the cardiac region only without truncating the heart.
Cine-CT free breathing protocols should be avoided unless the PET/CT system is designed by the manufacturer to perform this scan at radiation dose comparable to either ECG triggering or helical scanning with dose modulation.
One of the most significant sources of artifact in cardiac PET is misregistration of the transmission and emission datasets in space. When a patient moves between these two scans, the AC cannot be properly applied without adjusting for the patient movement (Fig. 4-14). This artifact is thought to be significant in as many as 40% of all PET/CT studies, thus if software is not available to align the images, a substantially lower specificity will occur.72 Imaging guidelines recommended that all cardiac PET studies be routinely examined for misregistration and corrected whenever possible.2
Misregistration artifacts are one of the most common artifacts in PET. These occur when the transmission and emission datasets are not aligned. When misregistration is present, technologists must correct images for these artifacts before proceeding.
By far the most common technique for misregistration correction is a rigid shift of the transmission and emission datasets.73 The user can interactively visualize an overlay of the transmission and emission data and move one of the datasets relative to the other until a satisfactory positioning is obtained. These offsets can then be used to re-reconstruct the tomographic data.
When the sinogram data are acquired in the 2D mode (septa extended), the scatter fraction defined as the ratio between the number of scatter events and the total number of coincidence events (scattered and unscattered) is from 10% to 15%, depending on the scanner geometry, the energy window, and the patient sizes.74 With septa retracted (no septa in the 3D mode), the scatter fraction can even increase to 30% to 50%, additionally depending on the maximum ring difference and span of the 3D data acquisition. Currently, most commercialized PET scanners offer options for scatter correction for the demand of imaging accuracy. For 2D PET imaging, implementation of nonstationary convolution–subtraction method has been used.74 The methods for 3D scatter correction are primarily relied on physical models for simulating the scatter process of the entire images.74–76 The 3D scatter simulation uses the activity images obtained from reconstructing uncorrected emission sinogram as the initial input of activity distribution to estimate the scatter component. The attenuation map functions to produce 3D scattering probabilities for each LOR of scatter events, based on the Klein–Nishina formula.77,78
Random events (randoms) are a result of unrelated photon events being recorded within the timing acceptance window of the system. Random events degrade image contrast, however because random photons are from unrelated annihilation events they only reduce overall image contrast. Hence, most randoms correction algorithm are implemented by subtracting a constant from the entire image.79
Protocols for Cardiac PET Perfusion Acquisition
The American Society of Nuclear Cardiology has developed imaging guidelines for the development of cardiac PET imaging protocols and should be consulted when developing clinical protocols.66 For a detailed explanation of protocols, see Chapter 10. Almost all cardiac PET perfusion studies are performed using vasodilator stress (although dobutamine may also be used). Patients should refrain from using of any product containing caffeine and other substance that could interfere with the vasodilator. For specific requirements, practitioners should refer the package insert of the vasodilator and national guidelines.80
For most studies, this single infusion site can be used for both the vasodilator and the 82Rb, particularly dipyridamole or regadenoson. A notable exception is when adenosine is used as the vasodilator. Because of the short half-life of adenosine, it is impractical to switch between the low flow rate adenosine to the high flow rate 82Rb without bolusing the adenosine into the heart or interrupting the flow of adenosine. Stress testing using 13N ammonia is considerably less restrictive than Rb-82, allowing for vasodilator and exercise testing.
Prior to the study, technologists will need to apply ECG patches for both the 12-lead cardiac monitoring system and 3-lead ECG triggering system for the scanner. Once patients have been prepared, they are positioned supine in the scanner system. Because AC is applied in all cardiac PET studies, it is possible to image patients with an arm down, however, it is important that the arm is immobilized and is free from the infusion activity.
The first stage of imaging is the transmission study, either acquired using CT or line source. For PET/CT systems, a low dosage chest planar image for positioning followed by a low dosage chest CT for AC will then be acquired of the mediastinal region. Can you explain how you use the map for positioning? For dedicated PET systems, a line source transmission study is acquired for both positioning and AC. How do you position? The transmission studies must be repeated if patient positioning is not correct.
The most common PET radiotracer for perfusion studies is Rubidium-82. Generally for this tracer, the resting perfusion study typically follows the transmission study. Delivery of the 82Rb using a generator cart requires strict adherence to the manufacturer's recommendations for dose delivery, QC, and general operations. Laboratories using an 82Rb generator should obtain training from the manufacturer prior to using the generator cart. To minimize the chance of patient motion between the emission and transmission study, it is important to begin the emission study as soon as practical. Most imaging protocols recommend a delay between the end of the infusion to the start of imaging of 90 to 120 seconds. Studies that measure flow however, require that the dynamic flow study be started prior to the infusion.81–83 The stress protocol immediately follows the rest acquisition.
These imaging studies can be acquired either using a list mode (inclusion of all photon pair events) or a frame mode. In principle, these two modes will yield similar results, however list mode acquisitions do offer some additional flexibility to create multiple studies from a single data acquisition. Practically, frame mode and list mode acquisitions do not appear to be different either quantitatively or visually.84
Because AC is applied prior to reconstruction, either FBP or iterative reconstruction can be used for creating 3D tomograms. Early PET systems utilized an FBP algorithm for reconstruction. More recently modern reconstruction algorithms for PET utilize an iterative, ordered subsets/expectation maximization algorithm because of the improved noise and image quality properties of the final reconstructed volumes.37,38
Iterative reconstruction algorithms use a stepwise updating algorithm to "search" for a source distribution that could produce the projection data observed. It relies on a model (the projector), of the transport of photons through the patient to the camera. In principle, this projector can be used to model any physical process in the photon transport, thus giving it considerably more flexibility than the FBP algorithm.
The iterative reconstructions PET in principle are similar to the reconstruction algorithms used in SPECT except the projection model is different. For a standard 2D acquisition, straight forward application of MLEM or OSEM is sufficient. If data are acquired in 3D, additional steps must be applied to take into account the oblique planes of data.
The simplest technique for reconstructing 3D data is to use the Fourier rebinning algorithm.85 This algorithm uses the geometric properties of the oblique planes to create a translation of the 3D data into traditional 2D planes.
Localization of the point of annihilation using the 511-keV photon pair to a line through the patient is improved upon by using "time-of-flight" (TOF) measurement capabilities.51 In this approach, additional information about the location of the annihilation event is obtained by measuring the (very small) difference in arrival times of the two coincidence photons. With conventional coincidence detection, the small difference in arrival times is ignored. TOF imaging requires very precise detector and electronic circuitry and calibration given the speed of light and the very short distances that the photons travel. Information provided by the difference in arrival times is included in reconstruction algorithms that model this modified probability information to improve spatial resolution and system sensitivity compared with conventional coincidence detection.86 It may also be used to improve separation of scattered photons from true coincidence events and system contrast resolution.