Introduction

The cardiologist’s arsenal of diagnostic tests and evaluations for patients with possible coronary artery disease (CAD) grew over the years [1,2,3]. As a result, our understanding expanded not only of CAD but also of the limitations of such tools. One of the best non-invasive imaging modalities that emerged has been positron emission tomography (PET). It has offered advantages over several other modalities in terms of diagnostic accuracy, image quality, and lower radiation exposure to patients [3, 4]. Furthermore, PET myocardial perfusion imaging (MPI) has become routine when PET is considered for evaluating patients with chest pain [5, 6]. The use of PET MPI has evolved further into a gold standard for quantifying myocardial blood flow (MBF) to evaluate myocardial viability and coronary microvascular artery disease (CMD) [6,7,8,9]. It has also become an excellent diagnostic test for certain inflammatory conditions of the heart, such as sarcoidosis and endocarditis, as well as for infections associated with implanted medical devices, such as pacemakers and implanted heart valves [10,11,12,13]. Certain technological improvements over the years have directly enhanced the utility of PET imaging as well as its image quality. Some of the developments were in the software used for the processing and analysis of images acquired by the PET scanners, while others were in the hardware of said scanners [14, 15]. One of the most recent improvements has been in the development of what is referred to as the “digital PET,” where the advent of silicon photomultipliers (SiPM) has offered certain advantages to replace the analog glass-enclosed photomultiplier tubes (PMTs) [16] (Fig. 1).

Fig. 1
figure 1

Digital PET detector with 3.2-mm LSO crystals silicon photomultiplier (SiPM) on the right; analog photomultiplier tube (PMT) detector on the left (Unrestricted© Siemens Healthcare GmbH, 2017 MI-3710)

Full size image

Earlier versions of the SiPM detectors were introduced to medical imaging applications in 2004. They had the advantage of the smaller size and compatibility with magnetic resonance, but without the time of flight (TOF) capabilities [17,18,19]. Conventional PET scanners that involved pulses generated by multiple single-photon avalanche diode (SPAD) arrays required off-chip processing and anger-logic positioning decoding. However, since SiPM offered the ability for one SPAD to detect one scintillation photon, there was no more need for off-chip processing and anger logic decoding [20, 21]. The development of digital PET has improved image resolution as well as the signal-to-noise ratio. It also has the potential to reduce the radiation exposure to patients since smaller doses of the radiopharmaceuticals would need to be injected, while maintaining excellent image quality. In this review, we describe the technological improvements in digital PET imaging as well as the advantages of using it in cardiovascular imaging. We also explore the pitfalls of using digital PET and the potential needed improvements. Finally, we summarize the available evidence on the prognostic utility of digital PET in cardiovascular imaging.

Technology Review: How Does Digital PET Work?

In PET/CT imaging, emitted gamma photos reach scintillation crystals, which in turn re-emit light photons that are directed to photomultiplier tubes (PMT) [22, 23]. PMTs are vacuum phototubes and are extremely sensitive detectors of light and multiply the current produced by incident light by as much as 100 million times producing voltage signals that are transformed into digital images using downstream electronic algorithms. PMTs have the advantage of high gain signal amplification, low noise, multichannel capabilities, and good timing performance. However, they have limited photon-to-electron quantum conversion efficiency and are relatively bulky.

Advances in PET/CT have resulted in the development of silicon photomultipliers (SiPMs) in lieu of PMTs [24, 25]. SiPMs are solid-state single-photon-sensitive devices that utilize single-photon avalanche diodes (SPAD) built on a common silicon substrate. The dimension of each single SPAD can vary from 10 to 100 µm, thereby decreasing detector size to 3 mm2 (compared to 6 mm2 with PMTs). In digital SiPMs, each SPAD is part of an array which operates as digital counter converting current signals to a digital bit of memory. Most digital PET scanners, however, use a hybrid (or analog) SiPMs where resistors are wired in parallel to get a readout of the total number of photons that are detected. These digital photon counters (DPC) have improved count rate sensitivity and can detect a dynamic range of photon detection down to a single scintillation photon. Current digital PET scanners use lutetium-based scintillation crystals directly coupled with individual SiPM DPC detectors (Table 1).

Table 1 Specifications comparison of three main digital PET scanners
Full size table

Compared to PMTs, SiPMs have improved light collection frequency with more efficient light collection. This results in improved energy and timing resolution. This also facilitates time of flight (TOF) measurements. During the process of positron annihilation, a pair of photons is released at 180 degrees opposite each other and reach the ring of detectors (coincidence detection). TOF measures not only the distance and attenuation of photons, but also the actual time difference (down to 6–200 picoseconds) between the detection of photons released during coincident events to more accurately identify the distance from the annihilation event to the detector. This translates to improved spatial resolution and better image quality. This also improves scanner sensitivity to detect ischemia, particularly subendocardial ischemia.

Clinical Applications

From a historical perspective, the SiPMs were initially developed to be coupled with hybrid PET/CMR scanners. This can explain one of their advantages, namely being resistant to magnetism [26,27,28]. Their most important advantages, however, extend beyond that, particularly from a clinical applications perspective. In particular, the smaller sized SiPMs allow having one detector for each SiPM, thereby eliminating the need for anger-logic positioning decoding and tremendously improving the count rate performance of the PET scanner, contributing to the enhanced spatial resolution [25, 29, 30]. Furthermore, digital PET offers a considerably improved TOF temporal resolution, which leads to enhanced effective sensitivity allowing an improved image quality, a reduced patient radiation dose, and a faster image acquisition time. Enhanced TOF also improves the convergence of iterative reconstruction, image contrast, and image quality in patients with high BMI.

  1. 1.

    Improved image quality and signal to noise

Digital PET systems provide better image quality than their analog counterparts [31,32,33]. Improving the pixel count in the image acquired by a PET scan can be achieved by extending the time interval of image acquisition, by increasing the dose of the radiopharmaceutical, and by enhancing the detection efficiency of the PET machine. Digital PET achieves a higher count rate by the third method, all while substantially reducing the imaging time and the radiopharmaceutical dose injected (Fig. 2). In fact, a major advantage of digital PET scanners is their ultra-sensitivity and rapid image acquisition times [34, 35]. This is because the major leap in digital PET was in the dramatically improved TOF timing resolution [36,37,38,39]. While conventional analog PET scanners take several [4,5,6,7] minutes to capture the required image depending on the radiopharmaceutical used, digital PET scanners only take a few seconds. As each microcell in the detector can be individually activated, data processing takes place within each pixel on the silicon chip leading to rapid and accurate ultra-low-light photon counting. This also allows for high counting rates without detector saturation, as well as fast response enabling sub-nanosecond timing (300–400 ps), both of which would be critical for first-pass image acquisition [33, 38, 40, 41].

Fig. 2
figure 2

Signal-to-noise ratio: a a digital PET image (5 mCi FDG), b a conventional 3DPET image (20 mCi) of the same patient

Full size image

The high count rate accuracy is particularly important for MBF quantification [42•, 43•]. Van Dijk et al. conducted a comparative study evaluating the myocardial blood flow quantification by digital and conventional PET scanners [43•]. They compared the count rate performance of the digital and conventional PET systems using the dynamic range defined as the maximal measured activity. They chose Rubidium-82 (Rb-82) for its short half-life (76 s) and compared the dynamic ranges yielded by the three different PET machines included (a digital PET prototype system from Philips, an Ingenuity TF from Philips, and a Discovery 690 from GE). They noted that absolute activity bias increased for all three scanners with increasing activity, but the digital PET prototype system suffered the least increase. Therefore, the count rate performance of the digital PET prototype was less affected at higher activities, and the applied dead time correction was considerably lower than in the conventional PET systems. They noted that this may present the opportunity for further corrections to improve the dynamic range of the digital PET scanner. Continued improvement in count rates as well as effective sensitivity would be expected to enhance the precision and reliability of MBF assessment by digital PET.

The background noise, defined as dark count rates, is processed within each individual counter effectively. The digital PET scanner operates by allowing each sensor to function independently. When only the sensors that have received sufficient photons above a certain threshold are activated, while those that did not meet the preconfigured threshold are deactivated, the nosiest cells (i.e., those with high dark count rates) are switched off [25]. The result is a much better signal-to-noise ratio.

  1. 2.

    Reduction in the injected radiopharmaceutical dose

A more efficient light photon collection method and higher effective time rates allows smaller crystals and lesser doses of the radiotracer to be used. Less ionizing radiation can also decrease the likelihood of crystal saturation, especially during the initial acquisition phases of the scanning protocol. The improved TOF with digital PET allows for yielding higher sensitivity with the ability to use a reduced dose of the isotope injected, as compared to conventional PET scanners [25]. One additional factor contributing to a lower dose of the radiopharmaceutical is the expanded axial field of view (FOV) of many of the digital PET scanners available for cardiovascular imaging [44]. An expanded axial FOV also reduces the time necessary for undergoing the imaging process [45]. The end results are lower stressor dose needed, lower patient radioisotope needed, lower elution volume, and a more efficient generator usage as it ages. All leads to a better patient experience as well as faster patient throughput.

Some labs with access to this technology have reduced their radiopharmaceutical injected dose significantly, while maintaining image acquisition time. For example, for Rubidium-82, 50–60 mci are injected when utilizing 2 cameras. This can be reduced to 25–30 mci with 3D systems and to 10–20 mci when utilizing digital PET scanners (Fig. 3).

Fig. 3
figure 3

A digital PET image of a patient with a BMI of 46 and a Rb-82 injection dose of 10 mCi. This was a 72-year-old female patient with DM type-2 who had presented with recurrent syncopal episodes, hypotension, and generalized weakness. Her cardiac PET scan showed a normal perfusion with a normal global myocardial blood flow reserve (2.2) and a coronary artery calcium score of 14

Full size image

In addition, detector saturation rarely occurs in these systems. Phantom experiments showed that saturation with Rubidium-82 occurs at 60 mci, a dose which is not used in clinical practice while utilizing these systems.

Potential Artifacts

As the sensitivity, TOF temporal resolution, and spatial resolution are higher for the SiPM PET scanner, the image quality becomes more sensitive to the artifacts, particularly those resulting from motion [42•]. The more apparent degrading influence of motion then requires even more need for correction. Motion blurring may lead to major artifacts in perfusion imaging. Artifacts stemming from motion tend to hamper image interpretation and may render PET scans as non-diagnostic, or worse, incorrectly interpreted. This emphasizes the need for continued advancement in multi-dimensional motion correction relating to both the heart movement as well as breathing [46]. Luckily, digital PET scanners enable data-driven motion correction where motion is ameliorated from the acquired raw images [47••] (Fig. 4).

Fig. 4
figure 4

Data-driven motion correction (DDMC): a before, b after motion correction

Full size image

Another potential artifact with digital PET stems from the improved spatial resolution as it exaggerates the appearance of apical thinning (Fig. 5).

Fig. 5
figure 5

A digital PET scan showing apical thinning. This was for a 64-year-old male patient with past medical history of CAD, hypertension, and hyperlipidemia. He had presented with chest pain. His cardiac PET scan showed a normal perfusion study, a borderline normal MBF (1.9), and apical thinning was noted on the digital PET scan

Full size image

Prognostic Value in Prior Studies

Few published studies have looked at the prognostic role of digital PET. Two prior studies were published from the Lausanne University Hospital in Switzerland. Dietz and colleagues assessed the prognostic role of digital silicon photomultiplier (SiPM) PET in 234 consecutive patients with suspected ischemia and followed over a median of 652 days for incident MACE (major adverse cardiovascular events: a composite of cardiac death, myocardial infarction, percutaneous coronary intervention (PCI)/coronary artery bypass graft (CABG) > 6 months post-imaging, hospitalization for heart failure or new onset stable angina) [48]. The authors evaluated global and minimum regional values for stress myocardial blood flow (MBF), myocardial flow reserve (MFR), and the myocardial flow capacity radius (MFC radius—the square root of the sum of stress MBF and MFR squared). This was a predominantly male (65%) older cohort (median age 72 years) with a high burden of known CAD and cardiovascular risk factors (54% known CAD, 73% hypertension, 36% diabetes, 68% dyslipidemia). A total of 47 participants experienced MACE during the follow-up period. Patients experiencing MACE had similar resting MBF, but significantly lower global and regional minimum stress MBF, MFR, and MFC radius. The receiver operator characteristics analysis found regional stress MBF of 1.94/1.7, MFR of 1.98/1.75, and MFC radius of 3.12/2.7 global/minimum regional values, respectively, achieved the ideal balance of sensitivity and specificity in identifying patients with MACE. These thresholds were significantly associated with the incident event in both univariable and multivariable cox regression models. In addition, the authors showed that MFC was incrementally prognostic over MFR and stress MBF.

More recently, the same group has also published an abstract at the Annual Congress of the European Association of Nuclear Medicine. Here, the authors aimed to compare the prognostic value of regional MFC to global stress MBF and MFR [49]. Although all 3 were independently associated with incident MACE in multivariable models adjusted for clinical factors, only stress MBF was associated with the outcome in multivariable models with all 3 metrics.

The largest study on the prognostic role of digital PET is that of Ahmed and Al Rifai et al. on 3678 patients with known or suspected CAD who were prospectively enrolled at the Houston Methodist DeBakey Heart and Vascular Center in Houston, Texas [50••]. Similar to the previous study, the cohort was composed of older patients (mean age 67 years, 52% men) with a burden of cardiovascular comorbidities and attendant known CAD (55% hypertension, 42% dyslipidemia, 32% diabetes, and 25% prior CAD). During a median follow-up of 8.5 (3.0–15.4) months, 229 patients experienced the primary outcome (all-cause death, myocardial infarction, and revascularization > 90 days after PET imaging). In multivariable models adjusting for traditional risk factors and relative PET perfusion variables, the authors showed that an impaired MFR (MFR < 2) was significantly associated with incident events. Furthermore, accounting for MFR improved model discrimination and risk reclassification. The aforementioned studies were all done using the Biograph Vision PET/CT scanner (Siemens Healthineers, Knoxville, TN, USA). Biograph Vision 600 has an axial FOV of 26.3 cm, a TOF performance of 214 ps, a sensitivity of 16, and an effective sensitivity of 79.1. MBF was also quantified using the one compartment model of rubidium kinetics as described by Lortie [51].

Conclusion

Overall, digital PET has been a great development in the technological advancements of non-invasive cardiovascular imaging modalities. It has offered superb utility for myocardial perfusion imaging and MBF quantification. Its high sensitivity and TOF allow lower activity and scan times, with much less risk of detector saturation. This allows faster patient throughput, scanning more patients per generator, and acquiring more consistent image quality across patients. The higher sensitivity captures more of the potential artifacts, particularly motion-related ones, which presents a challenge that still needs to be tackled. Data-driven motion correction, as well as machine learning motion correction, may prove helpful in that regard.

Key Points

  • Digital PET has been an important advancement in non-invasive cardiovascular imaging, particularly for quantification of myocardial perfusion imaging.

  • Silicon photomultipliers in digital PET have offered substantial improvements in several aspects of PET image acquirement, leading to higher sensitivity.

  • Further improvement in motion correction is still needed to optimize the quality of digital PET images.