Radiology
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
 QUICK SEARCH:   [advanced]


     


Published online before print March 24, 2004, 10.1148/radiol.2312021185
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Right arrow All Versions of this Article:
2312021185v1
231/2/305    most recent
Right arrow Submit a response
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Rohren, E. M.
Right arrow Articles by Coleman, R. E.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Rohren, E. M.
Right arrow Articles by Coleman, R. E.
(Radiology 2004;231:305-332.)
© RSNA, 2004


Special Review

Clinical Applications of PET in Oncology1

Eric M. Rohren, MD, PhD, Timothy G. Turkington, PhD and R. Edward Coleman, MD

1 From the Department of Radiology, Duke University Medical Center, Rm 1410, Duke North, Erwin Rd, Durham, NC 27710 (T.G.T., R.E.C.); and the Department of Radiology, Mayo Clinic, Rochester, Minn (E.M.R.). Received September 17, 2002; revision requested November 13; revision received April 14, 2003; accepted May 1. Address correspondence to R.E.C. (e-mail: colem010@mc.duke.edu).


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 TECHNICAL CONSIDERATIONS
 CLINICAL APPLICATIONS
 DEVELOPING APPLICATIONS OF PET...
 REFERENCES
 
Positron emission tomography (PET) provides metabolic information that has been documented to be useful in patient care. The properties of positron decay permit accurate imaging of the distribution of positron-emitting radiopharmaceuticals. The wide array of positron-emitting radiopharmaceuticals has been used to characterize multiple physiologic and pathologic states. PET is used for characterizing brain disorders such as Alzheimer disease and epilepsy and cardiac disorders such as coronary artery disease and myocardial viability. The neurologic and cardiac applications of PET are not covered in this review. The major utilization of PET clinically is in oncology and consists of imaging the distribution of fluorine 18 fluorodeoxyglucose (FDG). FDG, an analogue of glucose, accumulates in most tumors in a greater amount than it does in normal tissue. FDG PET is being used in diagnosis and follow-up of several malignancies, and the list of articles supporting its use continues to grow. In this review, the physics and instrumentation aspects of PET are described. Many of the clinical applications in oncology are mature and readily covered by third-party payers. Other applications are being used clinically but have not been as carefully evaluated in the literature, and these applications may not be covered by third-party payers. The developing applications of PET are included in this review.

© RSNA, 2004

Index terms: Breast neoplasms, PET, 00.12163 • Gastrointestinal tract, PET, 78.12163 • Head and neck neoplasms, PET, 26.12163, 27.12163 • Lung neoplasms, PET, 68.12163 • Lymphoma, PET, 99.12963 • Melanoma, **.121632 • Review


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 TECHNICAL CONSIDERATIONS
 CLINICAL APPLICATIONS
 DEVELOPING APPLICATIONS OF PET...
 REFERENCES
 
Positron emission tomography (PET) is a molecular imaging technique that provides images of physiologic processes. PET was developed in the early 1970s soon after x-ray computed tomography (CT) and at about the same time as magnetic resonance (MR) imaging (1).

The initial uses of PET were to help characterize disorders of the brain and heart by quantitatively imaging biologically important molecules labeled with short-lived positron-emitting radionuclides (Table 1). The radiopharmaceutical that has had the most impact on clinical PET imaging is 18F fluorodeoxyglucose (FDG), which was first described in the late 1970s (2). A chemist, Otto Warburg, had observed in the 1930s that malignant transformation of cells is associated with an increased glycolytic rate (3). FDG, an analogue of glucose, is metabolized similarly to glucose in that FDG is transported across cell membranes by glucose transporter proteins and is enzymatically phosphorylated. Once phosphorylated, FDG-6-phosphate is metabolically trapped, in contrast to glucose-6-phosphate. The primary exception to the metabolic trapping is in the liver, where the large concentration of phosphatase enzymes results in dephosphorylation of the FDG-6-phosphate and clearance of FDG from the liver.


View this table:
[in this window]
[in a new window]

 
TABLE 1. Radionuclides Used in PET

 
Because of the similarity of FDG and glucose, glucose metabolic rates can be quantified with PET by using appropriate modeling. Glucose metabolic rates are not used clinically. A semiquantitative index of glucose metabolism, the standardized uptake value (SUV), may be used for lesion characterization as a marker of glucose metabolism. The SUV is obtained by placing a region of interest over the lesion and dividing the value (in microcuries per cubic centimeter) by the injected dose (in microcuries) divided by the patient’s body weight (in grams). The software for calculating this index is available with the imaging systems.

Similar to CT and MR equipment, PET instrumentation has improved dramatically since its development. The system resolution for patient imaging units is now 4–5 mm, whereas it was greater than 15 mm for the initial systems. Animal imaging systems are now available with a resolution of approximately 2 mm. CT scanners are now being combined with PET scanners. The combined PET/CT devices offer several potential advantages overa PET scanner alone: better quality PET images because of the more accurate correction for attenuation provided by CT, automatic registration of CT (anatomic) and PET (metabolic) information, and shorter imaging times. The practical advantages and limitations of PET/CT scanners and the issue of cost versus benefit are currently topics of investigation.

The clinical applications of PET in oncology have been developed from the 1980s to the present time. The Health Care Financing Administration, which is now the Centers for Medicare and Medicaid Services (CMS), first approved PET for coverage in 1998. The first two indications were evaluation of the indeterminate solitary pulmonary nodule and initial staging of lung cancer. Coverage has subsequently expanded to include diagnosis, staging, and restaging of lung cancer, colorectal cancer, esophageal cancer, head and neck cancer, lymphoma, and melanoma. In addition, breast cancer is covered for staging and restaging and for therapeutic monitoring. Additional malignancies will be covered in the future.

This review will briefly cover the physics and instrumentation aspects of PET. We will focus on the clinical applications of PET and the relationships of findings from PET to findings from other imaging modalities. The limitations of PET will be addressed in the discussion of the malignancies.


    TECHNICAL CONSIDERATIONS
 TOP
 ABSTRACT
 INTRODUCTION
 TECHNICAL CONSIDERATIONS
 CLINICAL APPLICATIONS
 DEVELOPING APPLICATIONS OF PET...
 REFERENCES
 
Positron Physics
PET is based on the detection of the two annihilation photons produced when a positron is emitted from a radioisotope. PET radiotracers may be labeled with any positron-emitting radionuclide. The five radionuclides listed in Table 1 have all been used for PET. 18F is by far the most widely used PET radionuclide.

The decay of 18F is as follows: 18F -> 18O + e+ + {nu}. Each decay produces an oxygen 18 (18O) atom, a positron (e+), and a neutrino ({nu}). Only the positron is relevant for imaging, since the 18O isotope is given very little energy and the neutrino is very difficult to detect.

Positrons are the antimatter counterparts of electrons, having the same mass (511 keV/c2, where c is the speed of light) as and the opposite (positive) charge of the electron. Once produced, the positron leaves the decay site and gradually loses kinetic energy in the surrounding tissue, as would an electron. As this occurs, the positron imparts its energy in small increments through ionization and excitation of nearby atoms. The path length of the positron depends on its initial energy, which in turn depends on the particular radionuclide. The path length is less than 1 mm for 18F. Once most of its energy is lost, the positron eventually annihilates with a nearby electron: e+ + e -> {gamma} + {gamma}. Two photons ({gamma}) are produced, and there are two important features of this reaction (explained by conservation of momentum and energy): Both photons have an energy of 511 keV, and they leave the annihilation site in opposite directions.

Figure 1 illustrates the processes of positron transit, positron-electron annihilation, creation of two photons leaving in opposite directions, and the subsequent arrival of the photons at two detectors. Because the photons leave the annihilation site in opposite directions, it can be deduced that the annihilation was somewhere in the column connecting the detectors; otherwise, both detectors could not have been hit.



View larger version (7K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 1. Diagram shows positron annihilation resulting in back-to-back photons. Photons are detected in detectors on each side. The location of the annihilation is limited to the shaded column. e+ = positron, e– = electron, {gamma} = photon.

 
PET Imaging Systems
A simple PET imaging system is illustrated in Figure 2. A pair of photons is shown leaving an annihilation site and subsequently hitting two detectors in the ring surrounding the body. Simultaneous detection of two 511-keV photons in any two detectors in the ring indicates that an annihilation occurred somewhere in the column connecting those detectors. Such an event is referred to as a coincidence. Each coincidence is recorded, and the final raw data sample is a number for each detector pair (called a line of response), indicating the number of coincidences measured for that pair. These raw data represent projections of the distribution of radiotracer in the body, which are subsequently reconstructed into cross-sectional images by means of the traditional filtered-back-projection algorithm or, more recently, an iterative algorithm such as the ordered-subsets expectation-maximization algorithm (4), which provide images with improved noise characteristics.



View larger version (35K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 2. Schematic of a simple PET scanner detecting an event. The radiotracer molecule leading to this event must have been in the central gray area. The ring of rectangles represents the detectors.

 
Because PET does not rely on mechanical collimation to locate the origin of the emitted radiation, the efficiency for detecting emitted radiation is higher than that for single photon nuclear imaging, in which radiation is limited to specific trajectories before it impinges on the detector. Approximately 100 times more events are detected with PET per decay than with single photon imaging. Even so, the fraction of all emitted photon pairs that are actually directed toward detectors in the ring is small (<1%). When attenuation effects (scatter or absorption of emitted photons in the body) are added, the number of detected events is typically less than 0.1%. Improved image quality and/or decreased imaging time come with improvements in detection efficiency, which is accomplished by using multiple detector rings and with three-dimensional (3D) imaging (discussed later).

Even with low detection efficiency, PET imaging is the most sensitive way to detect small concentrations of tracers in the body. Typical tumors detected with PET have approximately picomolar concentrations of PET pharmaceuticals. Concentrations of CT and MR contrast agents must be many orders of magnitude greater to be detected.

Factors That Degrade PET Images
PET images are degraded by several physical factors. Background events include random events and scatter. Random events occur when two photons from unrelated annihilations are detected simultaneously and appear to be a true coincidence, as shown in Figure 3. The number of random events recorded is reduced by keeping the rates at which the individual detectors are being hit low and by using detectors that provide very precise timing.



View larger version (25K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 3. Schematic shows factors that degrade PET images. Top: Random event. Middle: Scattered events in transaxial plane (left) and out of plane (right). Bottom: Attenuation event, in which one of the photons never reaches its detector.

 
Scattered events are those in which one or both of the emitted photons undergo Compton scattering but are still detected. Most scattered photons would not be detected, due to loss of energy or to absorption in the body or because the new trajectory would be away from the detectors. The fraction of measured events that have scattered is larger for a larger body section; therefore, smaller patients yield lower scatter fractions, and brain imaging yields lower scatter than do most body sections. Scattered events can be rejected on the basis of their lower energy, but only if the detector has energy resolution good enough for a particular energy loss. Events that scatter from out of the axial field of view, as shown in Figure 3, can be rejected by the use of septa. These shields prevent radiation originating outside the field of view from impinging on the detectors, thereby preventing the detection of scattered coincidences. These septa improve imaging by limiting the count rates on the detectors, thereby minimizing random events and losses due to "dead time" (count-rate limits).

Both random and scattered events yield background that reduces the contrast of the resulting images. The real effect is that once the original contrast is restored (through corrections for scatter and random events or even through adjustment of the gray scale), the images appear as they would have if there had been fewer true counts.

Statistical noise is another degrading factor in PET, as in all imaging modalities. In PET, image noise is decreased if more true counts are acquired during the scan (as long as the increase in true counts is not accompanied by an even greater increase in background events). The number of counts acquired during a scan can be increased by either scanning longer or increasing the injected dose, although both have limits. Patient scan times are limited by how long a patient can rest without motion and by how many studies need to be performed on a scanner during the day. Dose is limited on some cameras by dead time. This is the case for systems with a few large detectors. Other systems, especially those with hundreds of independent detectors shielded by septa, can handle doses higher than are routinely prescribed.

Ultimately, the best improvements in counts come from improvements to the scanner that allow more counts to be detected for a fixed dose and scan time. One such improvement is to extend the axial field of view. For the task of whole-body scanning, more time can be spent at each table position if the scanner’s axial field of view is longer. A longer field of view generally implies more detector rings and more resultant transaxial image planes. Another improvement in detection efficiency comes by operating in 3D mode. In 3D mode, events are acquired regardless of any axial difference in the detected photon positions. In two-dimensional PET, in contrast, events are accepted only if the two photons are in the same or nearby rings. The use of septa (to minimize scattered events, random events, and dead-time losses) is only possible for a two-dimensional acquisition. The presence of the septa would preclude the additional events that 3D imaging tries to allow. A system that allows two-dimensional and 3D imaging must have septa that can be retracted. Of the currently available systems, some allow both two-dimensional and 3D imaging; the rest are 3D only.

Spatial Resolution
The intrinsic spatial resolution of a PET scanner is dictated by the ability to localize each measured photon. For systems with discrete detector elements, smaller elements generally yield better spatial resolution. Detector materials with high stopping power for 511-keV photons have the best potential for providing high spatial resolution.

The spatial resolution in reconstructed images is degraded in comparison to the intrinsic resolution of a system because of the noise-reduction filtering that is applied. Therefore, to obtain very-high-resolution clinical images, very-high-count scans are required to minimize the amount of smoothing necessary for acceptable image noise levels.

Attenuation
Attenuation, the loss of true events through scatter and absorption, is the largest source of qualitative and quantitative inaccuracies in PET imaging. Attenuation effects are worse for PET than for single photon imaging because both photons must survive passage through the body for the event to be counted, and therefore the combined paths for both photons form the relevant trajectory in calculating attenuation for an event. Approximately 10% of 511-keV photons are lost for each centimeter of tissue traversed. Several prominent artifacts appear on PET images of the body if no attenuation correction is performed. These artifacts include apparent high tracer uptake in low-attenuation regions such as the lungs, distortions of high-uptake areas such as the bladder (due to more attenuation in some directions than in others), and a prominent body edge that resembles high skin uptake. These effects are demonstrated in Figure 4, where a single section is shown with and without attenuation correction.



View larger version (42K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 4. Attenuation effects. PET emission images were obtained with (left) and without (right) attenuation correction. Artifacts on the noncorrected image (right) include the outer bright rim, artificially bright lungs, and elongation of the intense lesion.

 
In addition to the qualitative artifacts caused by attenuation, there are large quantitative inaccuracies if no attenuation correction is applied. The uptake of radiotracer in the body would be measured as only 5%–20% of the actual value without attenuation correction, depending on the body region.

The loss of counts due to attenuation leads to higher noise levels on images in larger patients. The noise implications of lost counts cannot be remedied by attenuation corrections, even though the qualitative artifacts and quantitative accuracy can be.

Attenuation Correction
Attenuation correction is accomplished by applying a correction factor to the number of measured events for each line of response. For example, if five events were recorded between a particular pair of detectors during a scan and it is known that only 10% of photon pairs emitted along that line would survive, than the corrected measurement is 5 · 10 = 50 counts. The determination of attenuation factors can be based on a calculation (assuming the outer contour of the body is known), an orbiting radionuclide source, or CT data. The calculation method is used primarily in brain imaging for those studies where high quantitative accuracy is not required.

Long-lived radionuclide sources have been used with conventional PET systems to measure attenuation in patients. Some systems have used germanium 68 (68Ge), a positron emitter with a 9-month half-life. With 68Ge, one or more sources orbit the body, and the detectors "look for" pairs of 511-keV annihilation photons, as in emission scanning. Other systems have used cesium 137, which has a 30-year half-life and emits 662-keV gamma rays.

A newer attenuation correction technique uses data from a CT system combined with the PET scanner. Although the attenuation properties of CT x rays are different from those of 511-keV emitted radiation, adjustments can be made to the attenuation map generated by the x rays to make it appropriate for calculating attenuation factors for 511-keV radiation for each line of response. In addition to providing attenuation correction, the superior anatomic detail of CT images, compared with the attenuation maps produced with radionuclide sources, is beneficial for aiding the interpretation of the matching PET images. An added benefit of the use of CT for attenuation correction is the acquisition time, which is faster than that for a rotating positron-emitting source.

Available Instrumentation
Consider two designs for PET detector components. In the first, the detectors are completely independent elements, each consisting of a scintillation crystal and a photomultiplier. A photon’s position is known from the location of the detector it hit. This is illustrated in Figure 2. In the second design there are few detectors, but each is a large gamma camera–like system that can measure the location of each detected photon through the processing of signals from an array of photomultipliers on the back (5). The presence of many independent detector elements gives the best count-rate performance, because the dead time of an incidence photon is limited to a very small detector area. Use of large-area detectors requires fewer photomultipliers, thereby lowering costs. Rotating gamma cameras modified for PET imaging are the lowest-cost example (6).

A blend of these two design types is the block detector (7). Block detectors are collections of small crystals (3–8-mm front-surface edge) that are grouped together into blocks of six by six or more. In some designs, the crystals are partially cut from a single block. In others, separate elements are put together. The light from each block is collected by a two-by-two array of photomultipliers. Even though the light from all (36 or more) crystal elements enters the same four photomultipliers, it is possible to determine which particular crystal was hit by comparing the pulse heights from the photomultipliers. A PET system using block detectors typically has tens of thousands of crystal elements arranged in hundreds of blocks (8,9). The block detector design allows much smaller crystals to be used than would be feasible if each crystal had its own photomultiplier.

Several different scintillators are being used in commercial PET systems. Desirable properties include good energy resolution (for rejection of scattered events), fast light output (for minimized dead time and rejection of random events), and high stopping power (for maximized detection efficiency and good spatial resolution). Rejection of random events, minimized dead time, and rejection of scattered events are especially important for successful 3D imaging. If the dose must be lowered for 3D scanning, then part of the benefit is lost.

The four scintillators currently being used in commercial PET scanners are shown in Table 2 (10), along with some important performance parameters. Stopping power increases with increasing density and increasing Z value (atomic number). Short decay times mean faster light output. High light output, in most cases, gives better energy resolution.


View this table:
[in this window]
[in a new window]

 
TABLE 2. Scintillators Used in PET Scanners

 
Each scintillator has advantages. Sodium iodide (NaI), which is slow and has low stopping power, has the best energy resolution. Bismuth germanate (BGO), which is slow and yields low energy resolution, has the best stopping power. Lutetium oxyorthosilicate (LSO) and gadolinium oxyorthosilicate (GSO), which are both fast, have stopping powers and energy resolutions between those of NaI and BGO. The cost of each material is a factor, as well, with NaI being the least expensive. PET systems are designed, ideally, to take advantage of the properties of the detector material being used. BGO systems are very good for the two-dimensional mode, where single rates, random events, and scattered events are controlled by means of septa. BGO detectors are also good for 3D imaging of the brain. LSO and GSO are good candidates for 3D-only systems, with the potential for very high count rates (11,12). Thallium-doped NaI, or NaI(Tl), is best suited for 3D imaging because of its energy resolution but is inherently limited by its low stopping power and therefore lower detection efficiency for 511-keV radiation.


    CLINICAL APPLICATIONS
 TOP
 ABSTRACT
 INTRODUCTION
 TECHNICAL CONSIDERATIONS
 CLINICAL APPLICATIONS
 DEVELOPING APPLICATIONS OF PET...
 REFERENCES
 
Non–Small Cell Lung Cancer
Lung cancer is currently the leading cause of cancer deaths among both men and women in the United States, with statistical estimates of 169,400 new diagnoses and 154,900 cancer deaths in 2002 (13). There are four main histologic types of lung cancer: squamous cell carcinoma, adenocarcinoma (including bronchoalveolar cell carcinoma), large cell carcinoma, and small cell carcinoma. All are thought to arise from pluripotent epithelial cells, with histologic features that depend on the pathway of differentiation. Within each group there are multiple tumor subtypes based on phenotypic and/or genotypic features. On a practical level, however, lung cancers are characterized as small cell or non–small cell in nature. Current CMS guidelines for reimbursement of FDG PET imaging is for non–small cell cancers only.

Diagnosis of Non–Small Cell Lung Cancer
One of the first widespread oncologic applications of FDG PET was in the characterization of pulmonary nodules, with the intent of estimating the likelihood of malignancy. Pulmonary nodules and other suspicious abnormalities are commonly encountered in clinical practice. Some are discovered during the work-up for chest complaints, while others are discovered incidentally during work-up for unrelated signs or symptoms. Pulmonary nodules are expected to be a common finding during proposed screening procedures for lung cancer (14).

The approach to the finding of a pulmonary nodule depends on a variety of factors. Some lesions can be demonstrated to be benign through certain distinctive radiologic features, including central, lamellar, or rim calcification (15). Lesions with a high likelihood of malignancy according to CT criteria, including invasion and adenopathy, may be addressed directly by means of either percutaneous biopsy or thoracoscopic surgery. Lesions amenable to percutaneous sampling in an otherwise healthy patient are often sampled directly with biopsy. However, many nodules and other pulmonary abnormalities are indeterminate according to imaging criteria, and either the patient or the referring clinician is hesitant about an invasive procedure and the associated risks and potential complications. In these cases, further imaging evaluation is appropriate.

Traditionally, a benign diagnosis for a pulmonary nodule is established through serial radiologic evaluation with either conventional radiography or CT (16). If the abnormality resolves, improves, or remains stable over time, it is considered to be a benign finding. However, benign nodules such as tuberculomas may enlarge over time, and some low-grade tumors may remain visually stable during the imaging evaluation period; therefore, there is debate over the issue of how much time is needed to establish a benign diagnosis (17).

The major drawback to the radiographic or CT approach is the time necessary to establish a diagnosis. At least one follow-up study, performed 3–6 months after initial demonstration of the abnormality, is needed to demonstrate growth and, therefore, the potential for malignancy. With this in mind, FDG PET has been shown to be a useful addition to CT in the evaluation of pulmonary nodules. Because of the preferential uptake of FDG in metabolically active tissues, PET can be used in a noninvasive assessment of the malignant potential of a pulmonary finding. Most non–small cell lung cancers accumulate FDG owing to a high metabolic rate (18) and are visually hyperintense on PET images (Fig 5). A majority of benign pulmonary lesions, including adenomas, hamartomas, and inflammatory or infectious nodules, do not accumulate substantial quantities of FDG and therefore appear hypointense on PET images (Fig 6) (19).



View larger version (127K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 5. Frontal (left) and left lateral (right) maximum intensity projections from FDG PET scan in an 80-year-old man referred for diagnosis of lung carcinoma. Newly discovered right lower lobe pulmonary nodule (arrows) is intensely hypermetabolic, with SUV of 12. Subsequent biopsy demonstrated non-small cell carcinoma.

 


View larger version (94K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 6a. FDG PET images in a 69-year-old man referred for diagnosis of lung carcinoma, with a newly-discovered left lung nodule. (a) Frontal maximum intensity projection is normal, with no hypermetabolic abnormality present. (b) Coronal sections from attenuation-corrected emission (left) and segmented transmission (right) scans. Transmission image demonstrates large nodule (arrow) in the posterior aspect of left lung. Only low-grade FDG activity is seen in this region on the emission scan (arrowhead). At patient’s request, the nodule was removed surgically and proved to be pulmonary hamartoma.

 


View larger version (103K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 6b. FDG PET images in a 69-year-old man referred for diagnosis of lung carcinoma, with a newly-discovered left lung nodule. (a) Frontal maximum intensity projection is normal, with no hypermetabolic abnormality present. (b) Coronal sections from attenuation-corrected emission (left) and segmented transmission (right) scans. Transmission image demonstrates large nodule (arrow) in the posterior aspect of left lung. Only low-grade FDG activity is seen in this region on the emission scan (arrowhead). At patient’s request, the nodule was removed surgically and proved to be pulmonary hamartoma.

 
A large body of literature exists that demonstrates the use of PET in distinguishing benign from malignant pulmonary nodules, and many studies have demonstrated a role for PET in determination of a benign or malignant cause for a pulmonary nodule (20,21). In a meta-analysis of the literature (22) encompassing 1,474 pulmonary nodules, PET was found to be 97% sensitive and 78% specific for malignancy. A semiquantitative means of estimating the likelihood of a benign or malignant cause is by use of the SUV (23). In a preliminary study on the average SUV of pulmonary lesions (24), investigators found that at a cutoff value of 2.5, the specificity of PET for a benign lesion was 100% and the sensitivity for a benign lesion was 89%.

Currently at our institution, nodule SUV is measured as an average pixel value within the lesion, calculated on the basis of body weight and measured on axial images. A circular region of interest with at least a 3-pixel diameter is placed over the most intense portion of the lesion. We consider nodules or masses with an SUV of less than 2.5 to have a very low likelihood, though not a zero likelihood, of being malignant. We consider nodules or masses with an SUV of greater than 2.5 to have a high likelihood of being malignant. Early studies indicate a possible role for dual-time-point PET for evaluation of these indeterminate nodules (25). For nodules less than 1.5 cm in size, the SUV criterion must be applied with caution. Although a nodule with an SUV of greater than 2.5 is always suspicious for malignancy, smaller nodules are progressively more difficult to evaluate accurately, and an SUV of less than 2.5 may be a reflection of a partial volume effect due to the limits of PET spatial resolution rather than of the underlying nature of the nodule.

With time, a better understanding of the benefits and limits of FDG PET imaging of pulmonary nodules has developed, and PET is now used to augment, rather than replace, conventional imaging strategies. Metabolic imaging with FDG PET has the advantage of requiring only a single point in time to perform an assessment rather than the multiple time points required for serial follow-up examinations. In this way, PET can expedite a more aggressive approach to a lesion that has a high likelihood of malignancy. Like serial radiographic or CT examination, false-positive results do occur with FDG PET imaging, primarily because of cellular and granulomatous inflammation such as can be seen in tuberculosis, fungal infections, and sarcoidosis (20,26). However, most hypermetabolic pulmonary nodules are malignant, and a lesion that is positive with FDG PET must, therefore, be assumed to be malignant until proved benign.

The high sensitivity of PET for malignancy means that non–FDG-avid lesions are highly unlikely to be malignant. False-negative PET scans are occasionally acquired, primarily in well-differentiated adenocarcinoma and bronchoalveolar cell carcinoma (23,27). For this reason, most patients with a negative PET scan are still followed up radiographically to firmly establish a benign diagnosis. PET-negative lesions followed in this way are usually truly benign, and those that do prove to be malignant are usually stage I nonmetastatic tumors, which have an excellent prognosis after resection (28).

Staging of Non–Small Cell Lung Cancer
Once the diagnosis of lung cancer has been established, the aim of evaluation turns to staging and treatment planning. The primary therapeutic modality for lung cancer is surgery, and complete surgical resection of disease offers the greatest chance for cure. Surgical cure is most likely in early-stage disease without nodal or distant metastases. Radiation therapy and chemotherapy are also effective in treatment of non–small cell lung cancer, either as independent regimens or as adjuvant therapy to surgical resection. The extent of tumor spread is the primary determining factor in whether surgical or nonsurgical therapy is offered to the patient.

With the widely used TNM staging scheme, classification of the primary tumor, T stage, is based on such factors as lesion size, presence of distal atelectasis, distance of the lesion from the carina, invasion of pleura or adjacent structures, and presence of a malignant pleural effusion. For most of these criteria, FDG PET is not sufficient to provide the necessary information for accurate staging. Although size measurements can be performed on PET images, such measurements include only metabolically active components of the tumor, and mismeasurement can result from incorrect windowing of images, with resultant visual "blooming" of the lesion and overestimation of lesion size. In addition, the resolution of PET and the delineation of margins are usually insufficient to enable determination of the presence or degree of invasion by the tumor, and such findings can only be questioned but not established with PET. Both lesion size and invasion are best determined with CT evaluation.

FDG PET is occasionally useful in the staging evaluation of the primary tumor, however. Central tumors can result in airway obstruction and distal atelectasis, which can occasionally have a masslike appearance. In such cases, CT may be inadequate to delineate tumor from obstructive or inflammatory change. FDG PET findings can augment the CT findings and may allow differentiation of metabolically active tumor from atelectasis or pneumonia. Likewise, CT and FDG PET are complementary with regard to the evaluation of intralobar satellite nodules, with size, number, and location demonstrated with CT and metabolic activity demonstrated with FDG PET.

The strongest role for PET in evaluation of T stage is in determination of the presence of malignant pleural effusion. Pleural effusions occur in patients with lung carcinoma, particularly if the tumor abuts the pleural surface or if there are postobstructive inflammatory changes. In many cases, these effusions are reactive in nature rather than a result of tumor seeding of the pleural space. CT, although sensitive for small quantities of pleural fluid, usually does not allow determination of whether the fluid is benign or malignant. The finding of increased FDG uptake that conforms to the pleural cavity on PET images has been found to allow differentiation of benign from malignant pleural effusions with an accuracy of 92% (29). A potential pitfall in this approach is in patients who have previously undergone pleurodesis, which can result in inflammatory pleural changes that are indistinguishable from pleural malignancy (30).

Nodal disease in non–small cell lung cancer is an important determinant of therapy and prognosis. Disease in ipsilateral hilar and mediastinal nodes or in subcarinal nodes is often amenable to surgical resection, and patients are usually still considered candidates for surgery. Nodal spread to the contralateral mediastinum or hilum imparts a worse prognosis, and these patients are usually treated nonsurgically. The most accurate means of establishing the presence of nodal disease prior to definitive surgical treatment is with lymph node sampling during mediastinoscopy. Although limited to superior mediastinal lymph nodes, this technique can usually provide sufficient information to designate an N2 or N3 status (31,32). The drawback of this procedure is its invasiveness, although in the appropriate hands it is associated with very low morbidity and mortality.

Noninvasive assessment of nodal disease is most often performed by using CT. Because of its dependence on anatomic characterization, CT evaluation of lymph nodes is based on node size and number in order to estimate the likelihood of malignancy. Nodes that measure larger than 1 cm in short-axis dimension are considered malignant, whereas nodes smaller than 1 cm in short-axis dimension are considered benign. In certain cases, numerous small lymph nodes are also considered to be suspicious. With this approach, CT is approximately 45% sensitive and 85% specific for metastatic disease to hilar and mediastinal lymph nodes (33). Modification of the size criterion may increase the sensitivity or specificity for disease but at the expense of the accompanying parameter: For example, a cutoff of 8 mm would increase sensitivity but decrease specificity.

The fundamental limitation of anatomic imaging is the inability to characterize the nature of lymph nodes. Nodes that are normal in size on anatomic images may contain tumor cells, and reactive lymph nodes can often exceed 1 cm in size yet contain no tumor. PET imaging offers metabolic information rather than anatomic measurements and is therefore an accurate means for delineation of nodal spread of lung cancer (Fig 7). When FDG PET has been used to characterize mediastinal nodes, sensitivities of 80%–90% and specificities of 85%–100% have been reported; for hilar nodes, the sensitivity and specificity of PET for metastatic disease are 75% (33). The lower specificity of FDG PET for disease in the hila is thought to be due to the frequent occurrence of inflammatory nodes in this region, which may be false-positive on FDG PET images.



View larger version (114K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 7. Frontal maximum intensity projection from FDG PET scan in a 75-year-old man with newly diagnosed non-small cell lung cancer, who was referred for initial staging of lung carcinoma, shows large metabolic mass in the middle portion of left lung (large arrow), with central photopenia suggestive of necrosis. Metastases to left hilar and mediastinal lymph nodes (large arrowheads), right adrenal gland (small arrow), and bilateral perinephric spaces (small arrowheads) are present.

 
FDG PET assessment of hilar and mediastinal lymph nodes can be used to categorize patients for surgical intervention. Negative predictive values for FDG PET evaluation of mediastinal lymph nodes are reported to be greater than 95%, and for this reason patients with no evidence of nodal spread on PET images can undergo surgical resection of the tumor with a high degree of confidence that the disease will prove to be limited at intraoperative assessment (34). With the potential for inflammatory disease to result in false-positive PET scans, the positive predictive value of FDG PET for mediastinal disease is lower, approximately 65% in some series (35,36). Because of this limitation, patients with metabolically active nodes shown on PET images should undergo a confirmatory procedure such as mediastinoscopy or transbronchial biopsy before an attempt at curative surgery is abandoned.

In many patients, it is the presence of distant metastases that precludes a surgical cure. A variety of sites can be involved by metastatic non–small cell lung cancer, with common sites of disease including lung, bone, brain, liver, and adrenal glands. Evaluation for metastatic disease with conventional imaging techniques requires several studies, including CT of the chest, abdomen, and pelvis; CT or MR imaging of the brain; and skeletal scintigraphy. FDG PET is an alternative means of evaluation for metastatic disease and in many ways is superior to conventional imaging techniques.

In one study of patients with non–small cell lung cancer and a negative staging CT scan (37), PET demonstrated unsuspected sites of disease in 11% of cases. In the same study, in patients with CT findings suspicious for metastatic disease, PET demonstrated nonmalignant causes in 95% of cases. FDG PET is also more accurate than skeletal scintigraphy with technetium agents for detection of osseous metastases (38). Benign adrenal nodules are a common CT finding, and in patients with lung carcinoma differentiation of benign from malignant causes of adrenal enlargement poses a challenge when standard imaging techniques are used. FDG PET has been shown to help reliably differentiate between adrenal metastases and benign adrenal nodules, with a sensitivity of 100% and a specificity of 80% (39).

A limitation of FDG PET in detection of metastatic disease is with regard to the brain. The cerebral cortex shows the highest physiologic tissue accumulation of FDG in the body; therefore, visualization of metastatic disease in the brain is problematic. The reported sensitivity of PET for detection of cerebral metastases is 85%, and the specificity is 73% (40). Small lesions are particularly difficult to discern with PET imaging; as a result, only approximately 65% of cerebral metastases are detected (40,41). For this reason, PET imaging of the brain cannot replace MR imaging for detection of cerebral metastatic disease.

Restaging of Non–Small Cell Lung Cancer
Despite aggressive therapy, the 5-year survival rate for patients with non–small cell lung cancer is only 14% (42). Some patients present with advanced disease at the time of diagnosis, but many patients complete therapeutic regimens intended to cure the disease only to have recurrence at a later date. Recurrent disease can arise in two ways: local intrathoracic recurrence and metastatic recurrence. FDG PET is useful for detection of both manifestations.

As discussed previously, FDG PET has been shown to be accurate in detection of metastatic disease at a variety of sites, and metastases to nodes, liver, bone, and adrenal glands are equally apparent in patients undergoing restaging of lung cancer and in patients undergoing initial staging. Similarly, new pulmonary nodules or masses in the ipsilateral or contralateral lung are accurately characterized with FDG PET.

An additional challenge in patients with previous lung carcinoma is differentiation of posttherapeutic change from recurrent tumor. Many patients with lung cancer undergo surgical resection—wedge resection, lobectomy, or pneumonectomy. Each of these leaves varying degrees of scarring and distortion that can mask early changes of recurrent tumor. With CT, serial examinations are required to document changes, and careful attention must be paid to each of the prior studies, lest subtle changes be overlooked. In contrast, PET imaging offers a rapid assessment of the status of disease and is accurate in differentiating postsurgical changes from recurrent tumor (Fig 8) (4345).



View larger version (89K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 8a. FDG PET images in a 62-year-old man with a history of non-small cell lung cancer. Patient had undergone right upper lobectomy and presented with increasing right upper extremity pain; he was referred for restaging of lung carcinoma. (a) Frontal maximum intensity projection demonstrates increased FDG uptake in apicomedial portion of right hemithorax (arrow). Normal laryngeal activity is seen in the midline of the neck (arrowhead). (b) Coronal image confirms hypermetabolic tumor adjacent to superior mediastinum and extending superiorly into the region of the brachial plexus on the right (arrow).

 


View larger version (84K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 8b. FDG PET images in a 62-year-old man with a history of non-small cell lung cancer. Patient had undergone right upper lobectomy and presented with increasing right upper extremity pain; he was referred for restaging of lung carcinoma. (a) Frontal maximum intensity projection demonstrates increased FDG uptake in apicomedial portion of right hemithorax (arrow). Normal laryngeal activity is seen in the midline of the neck (arrowhead). (b) Coronal image confirms hypermetabolic tumor adjacent to superior mediastinum and extending superiorly into the region of the brachial plexus on the right (arrow).

 
When radiation therapy is employed, pulmonary parenchymal changes can also result. Often, the CT findings of radiation-induced pneumonitis do not present a diagnostic dilemma, but on occasion the changes can be masslike and mistaken for tumor. In addition, when radiation therapy is employed as a primary treatment modality for lung cancer, the tumor mass may regress during therapy but will rarely resolve entirely. PET may be of use in these settings, and a negative PET scan makes recurrent tumor highly unlikely. However, radiation-induced pneumonitis is frequently seen to be metabolically active on FDG PET images (45), possibly owing to the cellular inflammation and macrophage response elicited by radiation-induced necrosis. Although this uptake is usually the most intense in the first 6 months after radiation therapy, cases of FDG uptake persisting for a year or more after discontinuation of radiation therapy are occasionally seen; therefore, the finding of increased FDG uptake inside the radiation port must be interpreted with caution.

Colorectal Carcinoma
Colorectal carcinoma is the third most frequent cancer in men and second most frequent cancer in women in the United States and was estimated to account for approximately 148,000 new cases in 2002 (13). Owing in part to screening efforts, many colorectal cancers are detected at an early stage, and the overall 5-year survival rate is 61%. However, up to 20% of patients will have metastatic disease at the time of diagnosis. Ninety-five percent of colon cancers are adenocarcinomas, the majority of which are moderately differentiated. Multiple histologic subtypes exist, with the subtypes based on the type of cellular differentiation noted at pathologic evaluation. CMS guidelines approve the use of FDG PET for diagnosis, staging, and restaging of colorectal cancer.

Diagnosis of Colorectal Carcinoma
In clinical practice, FDG PET is rarely used for the diagnosis of colorectal cancer, and state Medicare reimbursement guidelines often indicate that this is expected to be a rare use of PET. There are, however, scenarios where this indication could be expected to arise, such as in a patient with an imaging-depicted colonic abnormality that is suspected of representing colon cancer but for which a definitive diagnosis has not been made pathologically.

To date, there are no studies of which we are aware that prospectively examined FDG PET as a means of differentiating benign from malignant colonic abnormalities. Results of retrospective studies suggest, however, that PET may be of use in this role. When the scans of patients referred for evaluation of known colonic neoplasms were reviewed, PET was shown to be highly sensitive for identifying the site of the primary tumor (Fig 9), with sensitivities ranging from 90% to 100% (46,47). In these studies, the specificity of PET was much lower, ranging from 40% to 60%, likely because of physiologic bowel-wall accumulation of FDG, as well as uptake of FDG in inflammatory bowel conditions. Numerous reports have also described focal bowel accumulation of FDG due to benign colonic polyps (Fig 10). For these reasons, the evaluation of the patient with suspected colorectal carcinoma rarely includes FDG PET as an initial modality.



View larger version (79K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 9. Frontal maximum intensity projection from FDG PET scan in a 70-year-old woman with newly diagnosed adenocarcinoma of transverse colon who was referred for initial staging of colorectal cancer. Intense FDG uptake is shown (arrow) in left midabdomen, in the region of the patient’s known malignancy. No metastases are seen.

 


View larger version (121K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 10a. FDG PET images in a 56-year-old man with newly discovered pulmonary nodule and incidental discovery of colonic adenoma. (a) Frontal (left) and lateral (right) maximum intensity projections demonstrate nodule (arrows) in middle portion of right lung nodule to be hypermetabolic (SUV = 9.0). Subsequent resection demonstrated squamous cell carcinoma of the lung. Also noted is hypermetabolic (arrowheads) focus in right upper abdomen. (b) Axial image demonstrates that hypermetabolic (SUV = 6.5) lesion (arrow) is located anteriorly, possibly contiguous with transverse colon (arrowheads). Subsequent colonoscopy demonstrated 3-cm tubulovillous adenoma, with no evidence of malignancy.

 


View larger version (61K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 10b. FDG PET images in a 56-year-old man with newly discovered pulmonary nodule and incidental discovery of colonic adenoma. (a) Frontal (left) and lateral (right) maximum intensity projections demonstrate nodule (arrows) in middle portion of right lung nodule to be hypermetabolic (SUV = 9.0). Subsequent resection demonstrated squamous cell carcinoma of the lung. Also noted is hypermetabolic (arrowheads) focus in right upper abdomen. (b) Axial image demonstrates that hypermetabolic (SUV = 6.5) lesion (arrow) is located anteriorly, possibly contiguous with transverse colon (arrowheads). Subsequent colonoscopy demonstrated 3-cm tubulovillous adenoma, with no evidence of malignancy.

 
Staging of Colorectal Carcinoma
As with non–small cell lung cancer, surgery is the primary therapeutic modality in the treatment of colorectal cancer, with chemotherapy and radiation therapy used primarily as presurgical or postsurgical adjuvants. More than three-fourths of patients with colorectal cancer have disease limited to the bowel or to regional pericolic or mesenteric lymph nodes at the time of diagnosis (48), and most surgeries for colon cancer are, therefore, performed with the intent to cure. Surgery also plays a role in advanced disease because of the risks of obstruction, perforation, and hemorrhage that can occur with invasive tumors.

Because of the role of surgery in both limited and advanced disease, staging classifications for colorectal cancer have been developed on the basis of surgical and pathologic criteria. Several staging systems have been developed for colorectal cancer, including the Dukes system and the Jass system. Most recently, the standard TNM classification has been recommended by the American Joint Committee on Cancer. The goal of each classification system is to aid in determination of a prognosis for the patient and the most effective means of therapy.

The dependence of staging schemes on surgical and pathologic information limits the ability of imaging to provide the appropriate information before surgery. Because of its limited spatial resolution and poor delineation of anatomic boundaries, PET is particularly unsuited to providing the precise information needed for local and regional staging. In the TNM nomenclature, PET is of little use for the determination of T stage, where precise depth of invasion is the primary determinant. Only in cases of gross serosal penetration and invasion of adjacent structures might PET be accurate. Similarly, assignment of N stage requires numeric assessment of pericolic and mesenteric nodes. Regional lymph nodes in colorectal carcinoma are frequently small and lie in proximity to the primary tumor mass. In addition, pericolic nodes often contain small quantities of tumor cells, which are apparent only at the time of histopathologic evaluation. For these reasons, PET has been found to be only 29% sensitive for regional lymph node metastases from colorectal cancer (46). The presence of FDG in pericolic nodes is a good predictor of disease, however, and the specificity of PET for regional nodal metastases is 96%.

Disease spread beyond the regional pericolic or mesenteric lymph nodes is considered to be metastatic disease. Colorectal cancer may metastasize via the lymphatics and involve internal iliac nodes or retroperitoneal nodes, depending on the location of the primary tumor. Hematogenous spread of colorectal cancer almost always involves the lung or liver, and metastases to other sites without lung or liver involvement are rare. The strength of PET imaging is in depiction of these distant metastases, both nodal and extranodal (Fig 11).



View larger version (87K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 11. Frontal maximum intensity projection from FDG PET scan in a 59-year-old woman with recent surgical resection of colon carcinoma, referred for initial staging of colorectal cancer prior to additional therapy. Numerous hypermetabolic lesions in the liver (arrows) and retroperitoneum (large arrowheads) are shown. Findings are compatible with metastases. A subtle focus of activity in right pulmonary hilum (small arrowhead) was thought to be inflammatory.

 
Hepatic metastases from colorectal cancer are common and have in the past limited the therapeutic options available to the patient. With the improvement in surgical and noninvasive techniques, single or multiple hepatic metastases are now amenable to several therapies, including hepatic wedge resection, percutaneous or intraoperative radiofrequency ablation or cryotherapy, and arterial embolization (4951). The success of these procedures is highly dependent on accurate disease staging, in order to prevent unnecessary procedures in patients with concurrent disease elsewhere in the body.

CT is the principal modality for preoperative evaluation of patients with hepatic metastases, and intraoperative ultrasonography (US) is considered the imaging reference standard for identification of hepatic metastases. FDG PET has proved to be a valuable addition to the standard preoperative work-up of patients being considered for curative hepatic resection of colorectal metastases. Both PET and CT are accurate for identification and delineation of metastases to the liver, and in small case series with high prevalence of disease, PET has been shown to be superior to CT, with sensitivities and specificities as high as 100% for PET (5254). More recent studies suggest that PET is most accurate for identification of hepatic metastases greater than 1 cm in diameter but is limited in its ability to demonstrate lesions smaller than 1 cm (55,56), particularly when intraoperative US is used as a reference standard.

Despite its sensitivity and specificity for hepatic metastases, PET as a single modality does not provide sufficient information for adequate presurgical planning. In addition to delineation of the extent of disease, it is important to identify the precise location of metastases according to standard hepatic anatomy. Further, the position of lesions in relationship to vessels or to the gallbladder may preclude a safe surgical or ablative therapy. Since PET offers little information on anatomic boundaries, this needed information is not provided by PET alone.

Hepatic resection and ablative procedures are directed toward elimination of disease in the liver and do not address other sites that may harbor tumor cells. For these costly and invasive procedures to be effective, there should be no sites of disease that cannot be addressed with the procedure. Accurate imaging, therefore, is critical in these patients. The greatest impact of PET imaging in patients with hepatic metastases is in detection of these extrahepatic sites of disease that would preclude a curative procedure.

CT is the standard imaging modality for detection of extrahepatic disease prior to an attempt at a curative procedure. However, CT relies on anatomic parameters such as lymph node size and as such does not directly address the question of the presence of viable tumor cells. As a result, CT may result in an overestimate of the importance of an anatomic abnormality that is unrelated to malignancy or an underestimate of the importance of a finding because size criteria are not met. In a meta-analysis of the literature (57), PET was found to be more sensitive than CT for disease in a variety of sites, including liver, abdomen, retroperitoneum, pelvis, and other sites of disease. PET and CT were equally sensitive for detection of pulmonary metastases. Overall, PET was 97% sensitive and 76% specific for metastatic disease, whereas CT was 76% sensitive and 56% specific.

Several investigators have also examined the incremental value of PET as an addition to CT and found that PET offers information beyond that from CT and that this information often affects patient care. Investigators have found that when PET is added to CT in preoperative planning for patients with hepatic metastases, additional sites of extrahepatic disease are identified in 11%–23% of patients (57,58). This frequently leads to a change in therapeutic management from a localized directed approach to a more systemic approach with chemotherapy.

An important observation was made in one of these studies (58). In that study, in which 43 patients who were referred for hepatic resection were evaluated with CT and FDG PET, survival data after surgery were compared with established survival data in the literature. In the 43 patients, PET demonstrated 10 sites of disease that were not apparent on CT images. In six cases, this disease precluded a curative surgical procedure, and these patients were treated nonsurgically. By incorporating FDG PET into the preoperative planning protocol and eliminating from consideration those patients with unsuspected extrahepatic disease, the authors found that the Kaplan-Meier estimate of overall survival rate at 3 years for their patients was 77%. Authors of numerous studies in the literature of similar patients undergoing hepatic resection for metastatic disease without the use of FDG PET report a survival rate of approximately 40% (59,60). Although the number of patients in the study was not sufficient for generalization, the results suggest that the addition of PET imaging to presurgical evaluation may improve survival by appropriately eliminating ineffective surgery in patients with inoperable disease. A limitation also highlighted in that study was that despite the improvement in 3-year survival rate, the 3-year disease-free survival rate was only 40%, indicating the presence of occult disease undetected by any means, including FDG PET.

Restaging of Colorectal Cancer
After primary treatment of colorectal cancer, patients are at risk for the development of recurrent disease. Disease recurrence can occur in a variety of sites, most frequently the lung or liver. As during the initial staging of colorectal cancer, FDG PET is sensitive and specific for the presence of metastatic disease and is more accurate overall than CT. In a study comparing the sensitivity and specificity of CT, serum carcinoembryonic antigen levels, and FDG PET (61), PET was found to be more sensitive than both CT and carcinoembryonic antigen level for detection of disease recurrence, and the specificity of PET was equivalent to that of serum carcinoembryonic antigen level. In that study, the identification on PET images of additional sites of disease had the benefit of preventing unnecessary surgical intervention, with a consequent cost savings of approximately $3,000 per patient.

A particular challenge in patients with prior colorectal carcinoma is differentiation of the sequelae of prior therapy, including surgical scarring and radiation fibrosis, from disease recurrence. This is most problematic with distal tumors, where presacral scarring and pelvic changes are common. With conventional imaging, serial examinations are frequently required, and the diagnosis of disease recurrence may take 3–6 months or more before slowly developing changes are apparent.

With FDG PET, the presence of metabolic activity in the presacral space is indicative of tumor recurrence (Fig 12), while postsurgical change is not hypermetabolic. PET has been shown to be accurate for differentiation of benign from malignant presacral changes and to be superior to CT and MR imaging in this regard (53,62). PET has a further advantage in that only a single study is necessary to make this determination, rather than the serial studies required with conventional imaging.



View larger version (109K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 12. Frontal (left) and left lateral (right) maximum intensity projections from FDG PET scan in a 67-year-old man with history of rectal carcinoma who was referred for restaging of colorectal cancer. FDG PET was performed after low anterior resection. Locally recurrent disease in presacral space at the site of anastomosis (thick arrow), hepatic metastases involving right and left hepatic lobes (arrowheads), and solitary pulmonary metastasis (thin arrow) are visible.

 
Malignant Melanoma
Melanoma is a relatively uncommon tumor, with an estimated 53,600 new cases in 2002 (13). Once a rare tumor, the incidence of malignant melanoma is increasing and today is approximately double that in 1970. In its early stage, melanoma is curable by means of surgical excision, and up to 85% of patients are cured with treatment following diagnosis. However, the remaining 15% of patients present with locally advanced or metastatic disease at the time of diagnosis. The approach to treatment in these patients is multidisciplinary and may include surgery, chemotherapy, radiation therapy, and immunotherapy. CMS guidelines approve FDG PET for diagnosis, staging, and restaging of melanoma.

Diagnosis of Melanoma
We are aware of no studies to date that have examined the role of FDG PET in the diagnosis of melanoma. Since melanoma is usually a cutaneous lesion and amenable to diagnostic biopsy in the office, there is rarely a need for a noninvasive imaging study to aid in diagnosis. Furthermore, little is known of the potential for FDG uptake in benign pigmented lesions and atypical nevi. In a series of 95 patients with advanced-stage melanoma, Tyler and colleagues (63) reported a single false-positive PET scan in a patient who ultimately proved to have a benign cutaneous nevus. To our knowledge, no studies have prospectively compared FDG uptake in benign and malignant pigmented cutaneous lesions in a large number of patients.

Initial Staging of Melanoma
In the patient with newly diagnosed melanoma, there are several features that impart prognostic information, including depth of invasion of the primary lesion, spread to regional lymph nodes, and presence or absence of distant metastatic disease.

FDG PET has differing influence on each component of the TNM staging system for melanoma. For the T stage, depth of tumor invasion is the primary determinant, with invasion of more than 4 mm denoting the highest stage, T4. This depth is below the practical spatial resolution of current PET scanners, which is at best 5 mm for a dedicated full-ring scanner. Ulceration of the primary lesion is also a factor in determining T stage, a feature also not evaluated with PET scanning. These issues are moot, however, in that the assessment of the primary tumor is made at the time of punch biopsy or surgical excision, which almost always occurs before the patient is referred for PET scanning.

The next step in work-up of known melanoma is the determination of local-regional spread of disease. When the primary tumor is confined to the dermis, melanoma spread occurs through lymphatic channels to the regional lymph node basin. In the past, staging and treatment protocols have included full lymph node dissection of the likely nodal basin in order to identify sites of disease. This has largely been replaced by sentinel lymph node (SLN) mapping procedures, in which a small amount of microfiltered technetium-labeled sulfur colloid is injected intradermally around the primary tumor (64,65). The colloid particles are cleared by means of lymphatic drainage and are trapped in the lymph nodes that drain that site. These nodes can then be localized intraoperatively with a gamma probe. When combined with intraoperative injection of methylene blue dye and histopathologic evaluation, this procedure has been shown to be highly accurate for determining the presence of nodal disease. In one series of patients with melanoma on an extremity (66), the sensitivity of SLN mapping was 100% and the specificity was 97%.

Primary lesions located on an extremity typically spread in a predictable pattern to the nearest nodal basin, the axillary basin for upper extremities and the inguinal basin for lower extremities. With the use of SLN mapping, it has been found that a high proportion of head, neck, and trunk melanomas drain to unexpected nodal basins. Occasionally, SLN procedures may reveal drainage to two lymph node basins, and patients with disease in both nodal groups have been found to have a worse prognosis (67).

SLN mapping has been established as a safe and accurate procedure for determination of disease spread to regional lymph nodes. A large body of literature also incorporates the information obtained through SLN mapping with patient outcome and prognosis. When FDG PET has been compared with SLN mapping, mixed results have been reported. In one study of 17 patients with primary cutaneous melanoma (68), PET had an accuracy of 94%, as compared with findings from SLN mapping. However, that study appears to be the exception, because other studies (6971) have found that PET is relatively insensitive for early nodal disease. In a study of 89 patients with stage I, II, or III cutaneous melanoma (69), the sensitivity of PET was only 17% compared with that of SLN biopsy and surgical exploration. The specificity of PET was much higher (97%), and the presence of focal nodal activity is a good indicator of the presence of disease (Fig 13).



View larger version (85K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 13. Frontal maximum intensity projection from FDG PET scan in an 80-year-old man after recent excision of melanoma from the left ear. The patient was referred for initial staging of melanoma. Two discrete hypermetabolic foci (arrows) in left supraclavicular fossa, compatible with regional nodal metastases, are shown. At surgery, patient was found to have micrometastatic disease to a third node that was not visible at PET.

 
The limitation of PET for evaluation of early nodal metastases is primarily one of tumor volume. Although melanoma is typically highly FDG avid, early spread of disease to regional lymph nodes often involves small deposits of tumor cells. In SLN procedures, excised nodes are not only examined as frozen sections but are also sectioned thinly and stained with hematoxylin-eosin. Nodes that are negative with this method then undergo immunohistochemical staining for the presence of minute foci of metastatic disease. With this technique, nodal spread of extremely small numbers of melanoma cells can be detected. It has been shown that for PET to demonstrate disease in regional lymph nodes with 90% sensitivity, a 78-mm3 volume of tumor must be present (72); in the patients in that study, the median total basin tumor volume was 28.3 mm3.

Although limited in assessment of the primary tumor (T stage) and early spread to regional lymph nodes (N stage), PET has been shown to have a strong role in detection of metastatic disease. Melanoma can metastasize to a variety of sites, including distant nodal sites, lungs, liver, gastrointestinal tract, adrenal glands, and bone. Melanoma is also known for metastasizing to unusual sites such as spleen, gallbladder, and cutaneous or subcutaneous sites. There is no consensus on the most appropriate treatment for metastatic melanoma, but many surgeons pursue aggressive surgical excision of metastatic disease if only one or a few sites of disease are apparent. Precise identification of the location and number of metastatic lesions is, therefore, important for surgical planning.

Because of the variety of sites to which melanoma can metastasize, PET has been shown to be more accurate than CT for determination of the presence and extent of metastatic disease (Fig 14). PET has been found to be 92% sensitive and 90% specific for metastases (73). More compelling is the effect of PET scanning on patient care. When PET is added to CT in evaluation of patients undergoing initial staging of high-risk primary melanoma, a change in patient care occurs in up to 90% of cases because unsuspected sites of disease are detected (68).



View larger version (92K):
[in this window]
[in a new window]
[Download PPT slide]
 
Figure 14. Frontal maximum intensity projection from FDG PET scan in a 46-year-old woman with malignant melanoma of unknown primary who was referred for restaging of melanoma. Innumerable metastases are shown, including cutaneous or subcutaneous, nodal, hepatic, splenic, and pulmonary sites of disease.

 
Restaging of Melanoma
Melanoma recurrence can take place after years of clinical remission and may be clinically inapparent until advanced disease is present. FDG PET can be used for surveillance after treatment in patients with high-risk stage III and IV melanoma and has been shown to facilitate accurate identification of sites of disease that may otherwise not be apparent.

As with metastases found during initial staging, FDG PET is highly sensitive and specific for identification of disease at variety of sites, including organ, nodal, and cutaneous or subcutaneous locations (Fig 15). Many surgeons advocate aggressive surgical excision of metastatic foci as they develop, and the precise distribution of disease is therefore of paramount importance to achievement of curative surgery. In patients with known recurrence of melanoma, the addition of PET scanning to CT has been shown to aid in identification of additional unsuspected sites of disease and in alteration of treatment planning in up to 20% of cases (63,73).



View larger version (119K):
[in this window]
[in a new window]
[Downl