Robert L. Bridges, MD

In these first years of the 21st century, the power of PET is undeniable. PET has made a tremendous contribution in oncology, neurology, and cardiology and promises to expand our knowledge of molecular biology even into the very wall, cytoplasm, and nucleus of the cells. The promise of PET, however, and its implementation at the level of community medicine have been woefully stymied by a lemming-like diversion to a single mode of imaging, that of the hybrid PET/CT. Out of this has come a rigidity of opinion rather than the flexibility and adaptability required to expand the use of this needed modality.

As community medicine grapples with ever decreasing reimbursement, the industry now suggests that everyone buy two machines at the price of two and one-half machines instead of one machine (the PET scanner) and advanced software required to fuse the CT and MRI scans with the PET scan. For the increased price, one must tolerate more artifact, less precise and repeatable SUVs, and less precise assessment of critical areas of the body prone to significant variability in position, particularly the lung bases, pleural recesses, and the adjacent upper abdomen. 1 As the amount of prosthetic metal placed in human bodies increases, there will be larger areas with compromised image integrity if only CT level (“dental” or “diagnostic”) x-ray attenuation is used. Oral and intravenous contrast administration optimized for CT can degrade PET/CT imaging. 2

The entire argument for the hybrids hinges on a single mechanical linkage of a table spanning two machines and the hope that these two disparate modalities will coincide in space and time scanning a living and moving human being.

HISTORY

Historically, the PET/CT development is impressive. Developing the partial ring tomographic (PRT-1 and PRT-2 ) PET scanners, Nutt and Townsend recognized that these half-ring scanners had spare space that could house a CT scanner. The PRT series was designed to allow less expensive PET scanners to be manufactured. Over half the cost of the PET scanners of that time was for the crystals. Reducing the number of crystals significantly reduced the cost of the scanner. With the rotating crystal arrays opposite each other, the other half of the gantry was empty. In this space, a CT could provide a slice-to-slice exactly positioned transmission correction image. Hagashi et al formulated the reconstruction algorithm for this concept in the late 1980s. The space within the scanner unfortunately proved too small, and the final hybrid that emerged years later is the configuration we have today. Since it is a hybrid, it is not a true PET/CT scanner. The transmission image on a standard PET scan is a CT scan using a sealed energy source similar or identical to the 511 keV photons being captured. Germanium 68 sources decay to the daughter, gallium 68, which is a positron emitter. The source has to be replaced yearly. Cesium 137 has an energy of 662 keV and lasts decades. Both create transmission images that are immune to metal artifact encountered in the body. However, these transmission images eat up one third of the scan time. An x-ray CT scanner could provide a good approximation of the body’s density faster. The original intent was only to make a faster scanner. The irony is that out of an inexpensive scanner design emerged the complex hyper-expensive scanner of today. PET is heralded as the new contrast agent for CT, but the major stumbling block to that concept is that the contrast carries the bulk of the important information and the CT is along for the ride, occasionally (maybe 15% of the time) giving some needed directions, but often getting in the way if the interpreter takes short cuts.

WHO IS PILOT IN COMMAND?

PET imaging is the Lear jet of nuclear medicine. Many physicians are stepping into the proverbial cockpit with the confidence born in a trainer airplane, trusting the autopilot (PET/CT co-positional overlays) to fly the plane. Thus, we have radiologists banking on their familiarity with cross-sectional anatomy to help them with their “unclear medicine” experience, gained from reading bone scans and that weeklong or weekend course in PET. Meanwhile, nuclear medicine physicians rely on their functional imaging to buck up their weeklong or weekend course in cross-sectional anatomy. Further confusing the issue are the specialists who promote group reads. We have PET/CT with nuclear medicine and body imagers, PET/CT with nuclear medicine and head and neck imagers, or PET/CT with nuclear medicine and musculoskeletal imagers, and are not even beginning to contend with PET/MRI. Do the words “lost in translation” begin to mean something? This complexity has been visited before in the past. Egyptians had doctors who specialized solely on the left great toe or the right eye.

The power of PET has been proven through years of studies and thousands of patients. After every other test was performed, PET, not PET/CT, changed the course of patient care by 20% to 30%, less in some cases and more in others. PET/CT’s recent claim to fame is stating its improved diagnosis in 15% of cases by helping to better localize the area of abnormality. One has to first see the abnormality on the PET scan before localization with fusion can help.

Figure 1. New cystic area with history of astrocytoma, s/p radiation therapy. PET/MRI shows no recurrent lesion.

I suspect that many physicians are short-cutting interpretations, using the co-registered color pictures for the primary read. Co-registered nuclear medicine imaging is not new. Quantitative analysis of nuclear cardiac studies is an early example. Later, SPECT added three-axis review, better contrast and spatial resolution, and better analysis of wall function and perfusion. Would you care to guess how many physicians relied on the computer’s analysis without visual verification then and now? Color is the primary domain of ultrasound and nuclear medicine. In nuclear medicine, color is a dangerous tool. It is called “false color” for a reason. Relying on color creates false thresholds and can hinder general interpretation. What attracts most people to PET/CT? The co-registered color images! As one pathologist mentioned, it looks like an old 19th-century colorized black and white photograph. Most of the time it should be treated as one.

I can agree that localization helps at times, but I do not read just PET. If I can get the needed information from the PET and can save additional imaging and its costs to the patient, I will. If I can arrive at the best diagnosis with comparison side-by-side of outside films or with new studies, whether from CT, MRI, and US, I will. If I can import the outside study as DICOM or bring in our studies, I will fuse. We apply fusion for both (F-18) FDG and (F-18) NaF.

FUSION: REACHING CRITICAL MASS

Some may think the pure PET versus PET/CT battle is over and that only turf battles over who will read them remain; I believe a smoldering civil war is about to ignite.

The pursuit of precision imaging concentrating on repeatable patient positioning both spatial, between separate modalities, and temporal, between follow-up studies, in iso-alignment for PET and CT, will neutralize that simple advantage of PET/CT and pave the way to better imaging in both the hybrid and modular environment. Precision imaging brings PET/MRI, PET/PET, and even PET/CT/MRI fusion into focus.

What is precision imaging? Simply put, precision imaging is the ability to accurately and repeatedly reproduce exact patient position from one scanner to another scanner and from one examination to another. It maximizes the ability to fuse PET to CT or MRI. It also maximizes comparative longitudinal fusion of PET to PET and PET/CT to PET/CT, and PET/MRI to PET/MRI examinations.

Three basic areas of positioning include (from coarse to fine): 1) major musculoskeletal, 2) respiratory, and 3) organ shift. The challenge for modular co-registration and for longitudinal study-to-study fusion is to minimize positional differences.

The biggest movement is from one scanner to another. Uniformity of position requires repeatable alignment of the moveable portions of the body. Head to neck, neck to body, shoulders to upper torso, upper torso to lower torso, and torso to lower extremities are the main elements. In our practice, we measure these basic indices. We use the exact same cushions in PET, CT, and MRI (if possible for MRI). We also have carbon fiber flat beds for the CT and PET for radiotherapy planning.

Respiratory motion is the next greatest change, and here both hybrid and modular have exactly the same challenges. Respiration changes the position of the thorax and adjacent organs. While the thoracic spine stays fairly stable, the ribs, sternum, and shoulders move, and the diaphragm tightens. The hilar areas move, the heart moves, and the adjacent liver, spleen, kidneys, adrenal glands, and stomach move. The motion dampens out further into the abdomen and pelvis. The pleural surfaces in the recess can overlap, obscuring several centimeters of recesses if only the end-expiratory CT is reviewed.

Nonrespiratory organ shift can occur. This is highly variable and its significance is uncertain, but PET/CT should not be immune. Anxiety can halt peristalsis in the early stages of any examination, but with time that can subside and peristalsis resumes, possibly right in the middle of the PET imaging. Shift happens.

Our software technique independently parallels work presented at RSNA 2004 by Massachusetts General Hospital. 3 The CT protocols we use include at least one sequence with full breath for the recesses of the lower chest as part of the needed images, usually noncontrast, while the main sequence—usually the helical phase with dynamic imaging during the bolus—is taken at full end expiratory to match the PET scan. Delayed imaging can be done in either full inspiratory or expiratory phase. MRI fusion of the brain is usually with the FLAIR sequence. Head and neck MRI is at least two planes, one T1 and the other T2, unless the contrast images are more useful. MRI of the spine fused with PET bone is fused in sagittal and transaxial planes using an FDA-approved dedicated workstation utilizing both rigid and flexible fusion. At least one case benefited from PET/CT/MRI co-registration to best evaluate the liver and adjacent organs. Mixing and matching multiple modalities offers the best results.

COMPARISONS FALL SHORT

There is an Achilles heel to the bulk of previous studies comparing the PET/CT overlay from the hybrid and the studies from separate PET and CT scanners. On the one hand, the PET/CT studies follow standard imaging protocol. The CT is performed at end-expiratory breathing, which is the best compromise for the position of the diaphragm and adjacent organs in the chest and abdomen. The software fusion studies are, for the most part, the standard full breath CT we inherited from the days of slow scanners. Positioning of the patient was standard willy-nilly for the standalone CT scan. All of the variables that will absolutely ruin a PET/CT scan were kept for the CT scans, which are used for the software fusion portion of these comparative studies. This creates more than just an apples to oranges comparison, it creates a stacked deck.

Figure 2. A 57 year old with lung cancer with soft tissue and osseous metastases. Fusion with 3 week old outside CT discordant. Changes on present PET show mediastinal shift and increased activity in right chest. Patient brought back for CT, which was fused with PET scan. New huge right pleural effusion is seen. (Click the image for a larger version.)

One recent study compared PET/CT with software fusion. 4 Once the separate CT and PET studies were imported into the computer, the auto-rigid button was pushed and then the auto-morphing button was pushed. This was a two-keystroke acid test of a garden variety CT and a garden variety PET scan with all the attendant variables. No manual alignment was used since the authors felt that the software was advertised to do all of the alignment. Would it have been too much of a challenge, after time spent having technologists look up and import studies, to just spend 30 to 60 seconds visually pulling the images into course alignment? This represents the difference between hitting the target with an aimed arrow versus the Hiawatha approach of shooting an arrow into the air.

When I recently pointed out at an RSNA presentation that fusion of the PET with an iso-positioned CT with end-expiratory breath hold could yield very good results, the rebuttal was, “Why would you do that?” I stated that I already have a high-performance 16 channel CT scanner, a good PET scanner, and the most advanced software fusion in the world. I can swap out my 16 channel CT for a 64 channel CT or swap out the PET for the latest, greatest crystal. And—as we hear the other shoe drop—I can fuse with MRI.

THE NEXT ILLOGICAL STEP

So enamored with hybrid systems, some propose a hybrid PET/MRI. To work, it would have to be a PET/CT/MRI scanner, but even so, just how much compromise do we suffer? I understand that one PET/CT manufacturer reduced the resolution of their PET scanner from sub-5 mm to about 7 mm for the present PET/CT. The true resolution is for the total system. At 7 mm resolution, a patient can move quite a bit before the change is recognized, and at 7 mm, small lesions are more likely to be missed.

If PET/CT has problems with motion, does not pure PET have those same problems? The longer scan times mean patient body motion must occur, but since the pure PET obtains the emission and transmission images for the same cylinder of body, this motion is not additive for the whole scan. A patient with arms over the head can bring them down to their sides between table step ping. The PET/CT image is obtained in seconds for correction of body motion occurring over minutes during the PET scan. Body motion detracts from PET/CT in two ways. The most important is that body motion or misalignment degrades the attenuation correction. This can interfere with calculation of the all important SUV, the calibrated measurement of metabolism. Motion also causes image fusion misalignment. Disease can be misplaced and mismeasured. A possible solution is retrospective motion and attenuation correction for PET/CT, but that is not in the cards yet. The pure PET constantly sees motion for both portions of the scan, but in a form I liken to specular image correction: the additive collection of information tends to subtract out the motion problems. In astronomy, taking multiple pictures of a heavenly object through the earth’s atmospheric distortion can produce a picture that is sharper than a single picture.

In 1903, the Wright Brothers lifted off from Kitty Hawk with a heavier than air ship controlled by mechanical linkages. Later this year, Airbus will lift 800 passengers into the air by computers: Our most advanced aircraft must fly by computer. Software trumps hardware in the 21st century. PET/CT scanners can be bought with basic overlay software, but will require some software fusion. The advanced adaptive morphing software is “off the scanner table” for both PET and PET/CT

Widespread introduction and use of PET is beneficial to patient care. Adding pure PET improves care 20% to 30%. We are still too early into PET/CT to understand what it adds. While PET/CT is the darling of the industry and the “boombox” of nuclear medicine, the modular concept of integration of optimized PET, optimized CT, and optimized MRI for fusion is far from dead and, I believe, is the viable future. The supporters of hybrids may want to look over their shoulders. The “Mighty Morphing PET Scanner” is stirring.

One final comment: In the world of advanced computer imaging with 3D, virtual studies, and multi-modality fusion, may I make one recommendation? The best way to be prepared and keep agile in the virtual world is to go out, buy, and play some of the more advanced computer games. That way, after learning how to find and kill the bad guys, you can go to work and find and kill the really bad guys of disease. Be it modular imaging or hybrid, happy and successful hunting.

Robert L. Bridges, MD, is dual boarded in diagnostic radiology and nuclear medicine and has been in private practice in California and Alaska for more than 18 years. He participated in the installation of the University of California-Irvine’s first MRI and PET scanners in the mid 1980s. He holds patents in conjunction with the University of California.

References:

  1. Erdi YE, Nehmeh SA, Pan T, et al. The CT motion quantitation of lung lesions and its impact on PET-measured SUVs. J Nucl Med. 2004;45(8):1287-92.
  2. Cohade C, Osman M, Nakamoto Y, et al. Initial experience with oral contrast in PET/CT: phantom and clinical studies. J Nucl Med. 2003;44(3):412-6.
  3. Krishnasetty V, Aquino SL, Halpern EF, Fischman A. Comparison of alignment of computer registered datasets acquired from dual PET/CT scan versus independent PET and CT scans [abstract]. RSNA Radiology. 2004;233(3):373. Abstract 1212NM-p.
  4. Allen-Auerbach M, Quon A, Weber WA. Comparison between 2-deoxy-2-[18F]fluoro-D-Glucose positron emission tomography and positron emission tomography/computed tomography hardware fusion for staging of patients with lymphoma. Mol Imaging Biol. 2004;6(6):411-6.