|David W. Townsend, PhD|
Following the development of the first prototype PET/CT scanner 1 and the introduction of the first commercial design in 2001, adoption by the medical imaging community has been surprisingly rapid. While objective peer-reviewed, scientific studies that establish combined PET/CT as superior to either PET alone or PET and CT software fusion are only slowly appearing in the literature, 2 overall response from the early PET/CT adopters has been enthusiastic. Such enthusiasm appears well justified based on the improved diagnostic accuracy that is being reported due to the ability to precisely localize functional abnormalities. Also, routine availability of registered anatomy and function is increasing physician confidence in the interpretation of both the CT and PET studies. A combined study is generally more convenient for the patient and can be completed in significantly less time than a conventional PET-only examination.
In current PET/CT designs, the two scanners are physically separate with the CT positioned anterior to the PET. 3 A benefit of this minimal level of hardware integration is that either system can then take advantage independently of the advances in CT and PET technology. Coincidentally, in the 3 years since the introduction of PET/CT into the clinical arena, there have been significant advances in both CT and PET technology, advances that have in consequence become incorporated into the new PET/CT designs. In particular, the appearance of multislice CT first with two, currently with 16, and next with the recently announced 64 detectors per row has created, with each advance, a demand for incorporation of the top-of-the-line CT scanner into the latest PET/CT. PET imaging instrumentation has also seen significant progress with the introduction of new, faster scintillators, improved image reconstruction algorithms, and better correction techniques. Fast scintillators reduce system dead time and improve count rate performance even at low activity levels, and statistically based reconstruction algorithms offer better signal-to-noise and fewer artifacts related to noise propagation. One such fast scintillator, lutetium oxyorthosilicate (LSO), 4 also offers increased light output and hence better positioning accuracy and scatter rejection than bismuth germanate (BGO). The superior performance of the faster scintillators is a major advantage for 3D imaging where the high levels of scatter and randoms seriously compromise image quality when slower scintillators with low light output are used. The higher sensitivity of 3D PET imaging offers greater flexibility for selecting scan duration and the injected activity levels of the radioactive PET tracer, since with low dead time and high count-rate capability, LSO detectors can operate, for example, at injected activities of 20 mCi or even greater of fluorine 18 fluorodeoxyglucose (FDG). Compared to BGO, the better positioning accuracy due to the increased light output from LSO allows the block detector to be cut into smaller crystals, thus improving spatial resolution. With current state-of-the-art detectors, an intrinsic spatial resolution for PET of about 4 mm can be achieved, although clinically, especially in larger patients, such high spatial resolution may not always be attained. Nevertheless, as illustrated in Figure 1, FDG uptake can be demonstrated with good activity recovery in small structures such as a 3.5 mm malignant lung nodule.
The use of CT images to generate the appropriate attenuation correction factors for the PET emission data obviates the need for a separate PET transmission scan, thereby substantially reducing both noise in the attenuation correction factors and the total scan duration. The reduction in scan time to less than 20 minutes for a whole-body study is convenient for the patient and increases throughput. Challenges remain for imaging larger patients, although the superior performance of LSO detectors in 3D, the use of statistically based reconstruction algorithms, and better scatter correction techniques have all contributed to a significant improvement in image quality for these patients. Image quality can also be improved by increasing scan time.
The ability to scan melanoma patients from head to feet without repositioning is an example of the improved scanner operation that has occurred since the first designs became available in 2001. Many of the restrictions in the early designs have been overcome or eliminated, thus resulting in a more reliable and user-friendly operation. Different patient orientations and protocols are now fully supported, with increased parallelism and flexibility. Repeat PET scans can be acquired and reconstructed with attenuation correction factors from the same CT scan, and a shared database facilitates image reconstruction of one patient during acquisition of another patient. Considerable progress has been made toward the integration of the acquisition and data processing computers, somewhat in contrast to the lack of integration of the actual imaging hardware. Both CT and PET acquisitions can be controlled from the same computer screen, including an integrated QC procedure for the two imaging systems that minimizes the preparation and setup time prior to imaging the first patient of the day.
IMAGE QUALITY IMPROVEMENTS
Reconstructed PET image quality has improved, not only through the introduction of fully 3D statistical algorithms, but also with the use of larger image matrix sizes (up to 336 x 336, for example). In some PET/CT designs, the mismatch between the diameters of the CT reconstructed field-of-view (50 cm) and those of PET (60 cm) has been addressed by reconstructing the CT data over an extended field-of-view to more closely match that of the PET. While this procedure does not provide diagnostic quality CT at the periphery of the field-of-view, it does improve the accuracy of the CT-based attenuation correction factors. Some recently announced large-field CT scanners can now provide a matched field-of-view without requiring extended reconstruction.
|Figure 1. A 67-year-old male patient with squamous cell carcinoma of the lung referred for a PET/CT scan to restage the disease after treatment. The CT scan (a) identified a small, 3.5 mm nodule in the right lung (arrow). The PET/CT scan (b) demonstrated uptake of FDG with an SUV of 3.1 in the nodule (arrow). Also visualized was a 3 mm x 8 mm nodule with an SUV of 6.6. The scan protocol was: CT: 159 mAs, 120 kV, 5 mm slices at 0.75 mm; PET: 9.5 mCi of FDG injected, 109 min uptake time, 3 min per bed position acquisition.|
The overall management of the data has also been improved by a closer integration of the imaging system with PACS, ensuring a relatively seamless transfer of the PET/CT images into the hospital database, irrespective of the particular vendor. The availability of large off-line storage arrays has increased the data storage and archiving capacity and provided storage for CT raw data as required. The reliability and flexibility of the archiving system have been increased with the storage of CT and PET images on CD and, eventually, on DVD.
The advances in both the imaging technology and the integration of CT and PET have considerably improved the workflow associated with PET/CT scanners. The flexibility to reconstruct images during acquisition and the availability of the results within 2 to 3 minutes of scan completion facilitate the logistics of high patient throughput. With increasing evidence that a 90-minute uptake period for FDG may be preferable to a 60-minute period and with the higher level of injected PET radioactivity that can be tolerated by the faster scanners, an adequate number of shielded uptake rooms is required to sustain high patient throughput. The large shared database and, above all, enhanced system reliability are essential to maintain high throughput. With the replacement of the PET transmission scan by CT-based attenuation correction, scan times as short as 15 minutes or less can be achieved. 5 The large number of CT images generated by a multislice CT for each patient and the large number of patients that can potentially be scanned per day on a combined PET/CT represent a significant workload for the radiologist and nuclear medicine physician. Advanced software tools to review and analyze the images are essential, and there has been considerable progress since the introduction of the PET/CT to develop tools specifically to facilitate the review of fused image sets. The procedure implemented by early adopters involved assembling a team of radiologists and nuclear medicine physicians to read each study, a procedure that is clearly impractical in high throughput environments where a single reader who is dual-certified in radiology and PET is increasingly becoming the norm.
|Nathan C. Hall, MD, PhD|
Since the inception of the PET/CT concept in the early nineties, the targeted application has been to oncology imaging, specifically outside the brain. This is because CT rather than MR is still the modality of choice for whole-body anatomical imaging in oncology, and because software fusion works equally as well as hardware fusion for the brain. Although high-performance CT may not be required for oncology, this has not prevented the appearance of PET/CT designs incorporating 16-slice CT scanners, with plans to go to 64 slices and maybe even higher. However, the main impetus really driving the development of high-performance CT is cardiology where fast scanning is essential to limit the effects of motion. The availability of these designs opens up the possibility for cardiac PET/CT applications, as are currently being explored at a number of institutions. Fast CT also addresses to some extent the respiration mismatch that arises between CT and PET when CT is acquired with breath hold at full inspiration and PET with shallow, tidal breathing. The 10-second scan time for the thorax achieved with the 16-slice CT permits breath hold at partial inspiration more closely matching the average position of the thorax during the PET scan. In addition, the PET scan may be respiratory-gated to reduce the effect of motion, albeit at the expense of increased statistical noise in each of the PET gates. Experience with the use of respiratory gating for radiation therapy can also be applied to PET/CT, and the incorporation of functional images into therapy planning is progressing steadily.
CT-BASED ATTENUATION CORRECTION
|David W. Barker, MD|
The advantages of CT-based attenuation correction are short scan times and essentially noiseless attenuation correction factors. Apart from the respiration mismatch, issues also arise over the use of intravenous or oral contrast-enhanced CT images to generate attenuation correction factors. The concern is that contrast-enhanced CT pixels may be scaled incorrectly generating artifacts on the PET image that could be misinterpreted as focal uptake. For intravenous contrast, the CT images can provide a clue to regions of potential concern that, together with a review of the noncorrected PET images, will generally resolve any ambiguity. For oral contrast, procedures have been developed to identify areas of significant enhancement and to scale them accordingly with the appropriate tissue scale factor. 6 The emerging consensus is that contrast-enhanced CT scans provide accurate attenuation correction factors in all but a minority of cases where misinterpretation can be avoided by carefully reviewing both the CT images and the noncorrected PET images. Such a protocol provides standard-of-care CT images without exposing the patient to dual CT scans, with and without contrast. However, in some studies such as those involving serial scans, it may be desirable to limit the radiation exposure to the patient by acquiring a low-dose, noncontrast CT for attenuation correction and localization purposes only.
|Figure 2. A 54-year-old male with a 15 x 8 mm left lung cancer (small arrows) and left hilar and mediastinal involvement (large arrows) as demonstrated by (a) coronal PET and (b) coronal PET/CT images. The patient is 5’7 and weighs 142 pounds. CT parameters are: 164 mAs, 120 kV, 5 mm slices at 0.75 mm; PET scan parameters are: 10.2 mCi of FDG injected 100 min prior to scanning, 3 min per bed position acquisition.|
Another area that has received considerable attention, particularly in the United States, is the design of protocols optimized for large patients. Patient size is a major factor affecting image quality, for both CT and PET. Although higher levels of positron-emitting radioactivity such as FDG can be injected into larger patients, attenuation drastically reduces the number of true, unscattered counts while randoms and scattered coincidence rates increase. The result is a steady decrease in image quality and signal-to-noise as a function of increasing patient weight. To illustrate this effect, Figure 2 shows a coronal section of a 142 pound patient imaged for 3 minutes per bed position and Figure 3 shows a coronal section of a 224 pound patient imaged for 4 minutes per bed position. The PET image quality, although somewhat degraded in the heavier patient (Figure 3), is still of good diagnostic quality.
IMAGING IN 3D
|Figure 3. A 73-year-old female with a 14 x 12 mm left lung cancer (arrows) as demonstrated by (a) coronal PET and (b) coronal PET/CT images. The patient is 5’3 and weighs 224 pounds. CT parameters are: 200 mAs, 120 kV, 5 mm slices at 0.75 mm; PET scan parameters are: 8.3 mCi FDG injected 94 minutes prior to scanning, 4 min per bed position acquisition.|
All traditional multi-ring BGO scanners incorporate septa, consisting of annular shields of lead or tungsten positioned between the detector rings. Their purpose is to shield the detectors from the high levels of scattered and random coincidences such as encountered when imaging large patients. The septa also limit the scanners to 2D acquisition by eliminating the acquisition of wide-angle lines of coincidence. Multi-ring PET scanners with retractable septa that allowed both 2D and 3D acquisition were first introduced in the early nineties. While 3D methodology was generally successful for brain imaging, 2D operation was advocated for whole-body studies, particularly in large patients, in order to limit the scatter and randoms rates. However, since septa also significantly reduce the true coincidence rate, increased levels of positron-emitting radioactivity must be injected into patients for 2D imaging compared to 3D. The recent introduction of faster scintillators such as LSO with improved count rate capability, lower randoms rates, and better scatter rejection has resolved most of the difficulties associated with 3D imaging in a BGO scanner. Nevertheless, very large patients (> 300 pounds) are a challenge even for LSO scanners operating in 3D, and it is frequently necessary to increase the imaging time per bed position, even though this only partially compensates for the effect of weight. For fixed scan duration, it may make more efficient use of the total time to image for different time intervals in different bed positions based on the variable magnitude of the attenuation effect throughout the body.
The rapid acceptance of PET/CT, particularly within the United States, together with a demand for the highest performing CT and PET components, has significantly impacted the imaging technology market. Until recently, only top-of-the-line PET/CT scanners were available from the major vendors, generally at a cost in excess of $2 million. The progression from dual-slice CT to 16-detector rows and greater has occurred in only 3 years, accompanied each time by a demand to incorporate the current state-of-the-art CT into a PET/CT design. Consequently, the top-of-the-line designs now include CT scanners with a performance generally exceeding that required for routine oncology imaging. While these high-end designs offer exceptional image quality and patient throughput, the cost may limit the adoption of this technology by medical institutions with smaller budgets. In such institutions, reduced CT performance for lower cost may be a more attractive proposition, thus creating the demand for a mid-range PET/CT design at a cost closer to $1 million. A recently announced design from one vendor combines a rotating LSO PET scanner with a four-slice CT scanner and is one of the first to respond to the mid-range demand. However, high patient throughput remains an important consideration, and at current billing levels, institutions are keen to maximize revenues. Subject to certain conditions, the CT portion of the PET/CT scan can be billed provided accepted standard-of-care protocols are followed. For example, a patient with a recent prior diagnostic CT and no interval treatment or clinical status change should not be billed for an additional CT. Since CT-based attenuation correction has been found to be accurate, with few exceptions, even in the presence of intravenous or oral contrast, full clinical CT protocols can be implemented. Once a PET/CT program has become established, referring physicians can be educated to request directly, where appropriate, a combined PET/CT scan rather than a CT scan followed later by a PET/CT for staging or definitive diagnosis.
In less than 3 years, 80% of the PET market has migrated to PET/CT, with more than 400 such devices installed in medical institutions worldwide and a current growth rate of 150 to 200 units per year. One prediction is that more than 90% of the PET market will be PET/CT by 2005, even though the growth rate is expected to slow. For oncology, it is anticipated that essentially all PET imaging will eventually be PET/CT. The recent announcement of restricted PET reimbursement for Alzheimer’s disease may create a demand for a PET-only neuroimaging device, and the ultimate role for PET/CT in cardiology has still to be established. It is clear, however, that, irrespective of the current cost of the PET/CT technology, it is unlikely that there will be a return to the days of PET-only imaging in oncology. The benefits of accurate localization of functional abnormalities, increased confidence in reading both CT and PET, and the potential to substantially increase patient throughput ensure that PET/CT will become established as the imaging technology of choice for oncology. With the rapid advances in both CT and PET instrumentation, the cost of the current top-of-the-line scanners will inevitably decrease in future to a level where such advanced technology should become affordable to all medical institutions.
Acknowledgements: The authors thank Misty Long from the University of Tennessee, Knoxville, and Brian Smith for helpful discussions. Financial support for the PET/CT development is provided by NCI Grant CA 65856.
David W. Townsend, PhD, is professor in the Departments of Medicine and Radiology at the University of Tennessee in Knoxville and director of the Cancer Imaging and Tracer Development Program.
Nathan C. Hall, MD, PhD, is assistant professor of nuclear medicine in the Department of Radiology at the University of Tennessee, Knoxville.
David W. Barker, MD, is assistant professor of radiology in the Department of Radiology at the University of Tennessee, Knoxville.
- Beyer T, Townsend DW, Brun T, et al. A combined PET/CT scanner for clinical oncology. J Nucl Med. 2000;41:1369-1379.
- Lardinois D, Weber W, Hany TF, et al. Staging of non-small-cell lung cancer with integrated positron emission tomography and computed tomography. N Engl J Med. 2003;348:2500-2507.
- Townsend DW, Carney JPJ, Yap JT, Hall NC. PET/CT today and tomorrow. J Nucl Med. 2004;45(suppl 1):4S-14S.
- Melcher CL, Schweitzer JS. Cerium-doped lutetium oxyorthosilicate: a fast, efficient new scintillator. IEEE Trans Nucl Sci. 1992;39:502-505.
- Halpern B, Dahlbom M, Quon A, et al. Impact of patient weight and emission scan duration on PET/CT image quality and lesion detectability. J Nucl Med. 2004;45:797-801.
- Carney JP, Beyer T, Brasse D, Yap JT, Townsend DW. Clinical PET/CT scanning using oral CT contrast agents. J Nucl Med. 2002;43(5): 57P.