Raising the Value of PET: Technology, Reliability and Reporting Take it Up a Notch

When we look at the future of PET, we think of new biomarkers and applications. Is conventional FDG PET maxed out? Or is there room to improve now? No it is not, and yes there is, molecular imaging experts say. While the science will advance as more specific biomarkers come to market, the slow speed of drug development can’t stop the progress of PET. Imaging experts from across the globe believe there is room for PET to become more clinically powerful and more insightful for referring physicians within the current realm of applications. It’s time to take a closer look at how PET is building value, now.

For PET to get better, we start with improvements in access. Technologies must be accessible and available to hospitals large and small, academic and rural as well as mobile users. Novel technologies cannot remain restricted to large centers and lofty budgets. Imaging systems and software also need to be strong, smart and reliable, with exams reproducible in all sizes and types of facilities. Technology today must be better and accessible.

And like all good relationships, communication among physicians is key to improving the value of patient care. The written report—or its electronic counterpart—must facilitate effective and efficient communication among reading physicians and referring physicians to offer diagnoses and plan treatment strategies to improve patient outcomes.

“Medical imaging in oncology is progressively shifting from describing the disease from an anatomic point of view to metrics,” says Prof. Frédéric Courbon, MD, PhD, chief of nuclear medicine at Institut Univeristaire du Cancer de Toulouse in France and professor of nuclear medicine. “It is not just about describing what structures are being seen but to convert images in number as a surrogate of either prognostic or an indication of treatment efficacy. Molecular functional imaging seeks deeper insight from a microscopic point of view and have better insight into cell biology to use imaging as a biomarker. It is a personalized treatment. It is very much about quality and accuracy of cancer detection and analysis.”

It comes down to the patient, says Geoffrey Johnson, MD, PhD, consultant in nuclear medicine in thoracic radiology at Mayo Clinic in Rochester, Minn. “When we say there is no cancer we can see, we need the utmost confidence in imaging to support that statement. As technology increases signal and decreases noise, we are increasing the accuracy of those words our patients long to hear, ‘no evidence of disease.’ That is huge.”

The challenges

In the PET/CT imaging world, technology blends the hardware of PET and CT imaging with sophisticated software. While image quality continues to improve as better iterative reconstruction software is developed, intrinsic troubles remain in that the amount of PET tracer in a patient and the amount of imaging time are limited by practical considerations. This limits the number of detected positron events and makes image noise higher than desired. Because of this limitation, PET images are filtered after reconstruction to make them more interpretable. The goal is to find the image that optimizes some statistic of the acquired data—without contending with the associated image noise.

“With oncology, the emphasis on the PET side has been the detection of uptake,” says Eric Rohren, MD, section chief of PET/CT and professor of nuclear medicine and radiology, MD Anderson Cancer Center in Houston. “Over the years, there have been improvements in scanner technology aimed at improving the resolution and count recovery of PET/CT to see very subtle activity and differentiate that activity from physiologic background. The scanners have definitely gotten better at seeing small lesions and detecting subtle sites of activity.”

Yet, variability of measurements in PET is a limiting factor. “The actual SUVs [standardized uptake values] we measure are not the actual SUV of the lesion,” Rohren notes. “They are the SUV compounded by spatial resolution, partial volume averaging and patient-specific factors we cannot control.”

Over the last decade, PET image reconstruction technology has been designed to provide better image quality, reduced acquisition time and lower injected dose. Current PET iterative reconstruction technologies, such as time of flight (TOF) and ordered-subsets expectation maximization (OSEM), force a compromise between image quality and quantitation. To improve the value of PET in therapy or response assessment, we need to improve the reliability and minimize the variability of SUV measurements with algorithms scan to scan.

Inside the technology

One way to limit noise in the PET/CT reconstruction is to stop the algorithm before the noise increases so much that it degrades the image. If the algorithm is stopped before convergence, it is important that the image has converged uniformly. To date, this has not been the case. In currently used PET/CT systems, large volumes of tracer converge more rapidly than small volumes. An organ, for example, converges before a small tumor within it. While uniformity of convergence has been improved with algorithms, no solution has led to full or nearly full convergence. Thus, PET images display quantitative artifact and may not provide an accurate SUV for all lesions.

GE Healthcare has rolled out an approach that changes the paradigm by augmenting the iterative algorithm so it converges to less noisy images. Called Q.Clear, this image reconstruction software is available on all GE PET/CT systems (there’s that access again), eliminating trade-offs between image quality and quantitative SUV measurements. The significant challenge with delivering consistently accurate SUV measurements in PET is that lesion size, volume and contrast recovery are highly impacted by today’s reconstruction techniques. Q.Clear is a significant step forward, generating images with low noise compared to clinically widespread OSEM while increasing quantitation accuracy due to full convergence, more rigorous noise control and edge preservation. This means the SUVs delivered, Q.SUVs, are fully converged, more accurate and more consistent. It can run the reconstruction to its full convergence, up to 25 or 30 iterations, without the penalty of noise increase. Research shows1 that by providing up to two times improvement in both quantitative accuracy (SUV mean) and image quality (SNR) in PET/CT imaging, this tool can provide bene.fits to physicians across the cancer care continuum from diagnosis and staging to treatment planning and assessment.

“In our experience, this method is able to suppress noise while continuing to do multiple iterations,” says Andrei Iagaru, MD, co-chief of nuclear medicine at Stanford University Medical Center and associate professor of radiology and nuclear medicine at Stanford University. Stanford has been using the new reconstruction algorithm for 15 months in research projects. Clinical scans on its DiscoveryTM PET/CT 710 scanner started in June 2015.

“They come very close to what the actual object should measure,” he says. “This is very important for quantitation as you want to be as close as in reality to what that measurement should be and you want to do it without reducing image quality.”

Iagaru and his colleagues have verified measurements independently, both in phantom and clinical data. “Based on our results, we plan to implement the new reconstruction algorithm in routine clinical practice.”

Q.Clear drives the reconstruction to full convergence. It does not solve all problems of quantitation. But at least it significantly minimizes the effect of under convergence. A PET report with multiple “lesions of interest” will be more consistent and reliable.

“If you improve the quality of the system and reduce the noise, you remove some of the reasons of uncertainty. To improve the signal to noise ratio, you will for sure improve the reproducibility,” says Courbon.

Iagaru says his group has found the technology very helpful in finding “very small lesions where we are really able to make them more conspicuous. And for all lesions, you are able to get the actual measurement closer to what they are in reality. In some instances, we see things that we don’t see with the [current] standard of care.”

A group at Mayo Clinic, that presented clinical data at SNMMI 2015, found that PET with Q.Clear can detect small foci of cancer with more sensitivity than without Q.Clear. “That is important because you can detect smaller sites of cancer,” says Johnson who was the senior author. “We were able to increase the signal and reduce the background [noise] with Q.Clear compared to standard time of flight PET, and we were able to see more tiny foci of cancer. We found the numbers were just shy of 50 percent in terms of the increase in signal to noise [in certain types of cancer]. That is critically important for our ability to confidently tell a patient he or she does or do not have evidence of cancer.”

Johnson also sees great potential for PET as we continue to see a shift to more minimally invasive surgeries. “We want better, more sensitive imaging. Advances in detecting metastatic cancers [in certain types of cancer with PET/CT or PET/MR] for example, may allow physicians to do less invasive surgeries to the benefit of patients.”  

Johnson stresses that his team, aided by Q.Clear, is more confident both when they can see potential cancer lesions that they are not fooled by noise, and more confident when no lesions are seen that there is no disease. “When we are assessing whether a patient should have surgery for lung cancer, we need maximum sensitivity. With a higher degree of confidence, this allows us to say, ‘no, I don’t think there is any cancer beyond what is appropriate to remove surgically.’ Of all the things that Q.Clear does for clinical practice, increasing confidence that there is or is no cancer present in a given location is the most important. This changes clinical practice and patient management.”

Improving reliability

“At the end of the day, we need to be sure the number has a strong relationship with the pathological pathway needing to be blocked by a treatment,” says Courbon. “PET is a good methodology to assess response to biotherapy. The cornerstone is quality assurance and reproducibility.”

With new technology that improves the signal to noise ratio and thus contrast, physicians may identify and qualify smaller lesions with confidence. The objective is for Q.SUVs to offer referring physicians more clarity and conclusiveness. Added to reports, physicians may garner more confidence in decision making. More accurate and consistent SUV measurements will make the PET picture more equipped to track the more complex types of disease progression.

Monitoring cancer treatment is not as simple as saying it is “getting better,” “getting worse” or “stable.” In most cases, the response is mixed—with some lesions improving and others getting more active. The art of monitoring cancer treatment with PET is closely related to following not one, but several strategically chosen lesions. If conventional imaging reconstruction induces a different level of errors depending on lesion size and location, monitoring SUVs in multiple lesions becomes a real challenge.

The judgment call is deciding if a particular lesion is benign or malignant. But ultimately it is understanding what impact that has on patient management. “My life gets easier as I start to understand how various tumors are treated and how the patient is managed and guiding appropriate therapy,” Rohren says. “It is a moving target because therapy is always changing. This requires close participation with referring clinicians and participating in tumor boards and really what they need from the PET. This comes around to reporting. The ultimate goal is to answer their question. Sometimes it is overt and sometimes not really stated. We need to understand what information is important for them to know what treatment this patient should get.”

Rohren says it is the professional responsibility of the physician to take it to the next level. “You can get it as close as you can with the technology but the physician needs to take it the remainder of the way. It comes down to your experience with PET, your comfort with the technology and sometimes just having seen enough scans to make those judgment calls.”

Building a better report

So that brings us to reporting, step three in improving the value of PET. The report needs to contain, hold and communicate more value. Its language and contents must continually improve so that reading physicians perceive PET as more conclusive and less subjective.

Improving the communication among physicians on care teams comes down to improving the language of the PET report. “The more standardization the better, for the patient, the physician and the referring physician,” says Iagaru, “in terms of histories, details and techniques.”

“We need consistency with accepted PET report templates,” echoes Rohren who was one of the authors of SNMMI’s 2013 Reporting Guidance for Oncologic 18F-FDG PET/CT Imaging that defines the essential elements of the PET/CT report. “It comes down to language and interpretation of numbers. Ultimately, it is about understanding the clinical questions and designing reports that answer questions for a specific scenario.”

Physicians are taking note and changing their practice, although the experts agree more facilities need to dedicate more time building better, more standardized and information rich reports. There needs to be more consistency based on format, content and quality. “Despite the guidelines, many are not using them,” says oncologist Anthony Shields, MD, PhD, associate center director for clinical services and program leader for molecular imaging at Karmanos Cancer Institute and professor of oncology at Wayne State University.

As the SNMMI guidance notes: “Generating a high-quality PET/CT report is perhaps more challenging than generating a report for other imaging studies because of the complexity of this hybrid imaging modality.”2 The content of the report not only influences patient management and clinical outcomes but also serves as legal documentation of services provided and can be used to justify medical necessity, billing accuracy and regulatory compliance.

Reports should include five elements: clinical history, technique and procedure, comparison studies, findings and impression. Each element includes a series of details. Clinical studies demonstrate that the availability of clinical history increases the accuracy of radiologic image interpretation. Referring physicians play a key role too in providing an appropriate indication for the study and clearly stating the primary clinical questions to be answered by the scan.

Another rule of thumb: understand the clinical situation you are imaging the best you can. “Scan clinic notes for insight,” Shields offers. “Is it a new 3mm lesion in the lung but the liver is stable? That is the patient’s life-limiting disease.”

The impression is the most important section of the imaging report, with many referring physicians starting there first. They read the findings section only as time allows. “Oftentimes oncologists look at the report and not at the images,” Shields adds. “The whole picture includes both.”

Overall, the guidance states “it is essential that all important information is presented in the impression in a clear and succinct way. The impression section should be a brief and concise interpretation of findings, not simply a restatement of findings. Lengthy discussion should be avoided in the impression.” The language used should be as clear and unambiguous as possible.

The impression should allow the reader to identify findings that are normal or abnormal, and answer the clinical questions raised by the referring physician. It also needs to include a clear-diagnosis or a brief list of differential diagnoses with a level of likelihood. In post-therapy follow-up studies, the metabolic response and anatomic response may be reported in the impression, especially if they are discordant.

Urgent or emergent findings—such as pneumothorax, impending pathologic fracture, spinal cord compression, or intracranial hemorrhage—should be communicated rapidly to referring physicians or their surrogate, and the date, time and means of communication should be documented at the end of the imaging report.

“The best way to improve the lines of communication with radiologists is to look at the report together or while on the phone, along with the images, look at series and slices, put the cursors on the lesion for a size measurement,” Shields says. “Before and after images are key to assessing treatment over time. Do you think it is better or worse? Or is it somewhat uncertain? Sometimes we modify the report based on the conversation when together we change our interpretation.”

And don’t forget the patient; patients read reports. “Even with a small lesion, we need to qualify that for the patient,” he says. “We can’t say this could be cancer. Is it likely or possible? You can frighten the patient when something is likely benign. That is not good reporting.” Avoid emotional terminology, as the guidance recommends, which is not helpful and often provokes unnecessary patient anxiety.

Brevity is key for the team at Mayo Clinic, Johnson notes. “We try to make sure that we A. answer the clinical question and B. put forth the imaging findings in a context that is useful to the clinician—while being very brief about it. Automatically adding detail and data does not excite me unless I am convinced it clarifies something or changes clinical decision-making.”

Depth and understanding

Beyond the reporting guidelines are the quantitative numbers and subjective opinions that matter most to patient management. Quantitation is the communication language between reading physicians and referring physicians. Improving quantitation improves communication. The art of monitoring cancer treatment is following several strategically chosen lesions. A more confident, more clear and user- friendly report is naturally more actionable. If oncologists get more insightful guidance, they will turn more often to PET.

“Quantitation helps very much,” says Courbon. “It is up to the referring physician to make the decision—carry on with toxicity or give up that there is no other solution. Sometimes it is very obvious, there are many lesions, the lesions have increased or just the opposite. But that is not for the majority of our patients. This is the gray area.”

Experts believe Q.Clear could improve the reading experience. With higher signal to noise ratio and more consistent quantitation, reading physicians could navigate faster around suspicious areas. This gain in efficiency would not only save time, but also enable the reader to focus on a more careful evaluation of the real areas of concern. As a consequence, they will be more empowered to call more relevant findings, with more confidence. A confident read adds a lot of value to the overall perception of the PET scan as a diagnostic tool.

“In the imaging community, we need to adopt these standards—in reporting and quantitation—so that we can take results from one center and accurately interpret it at another center,” Iagaru notes.

Improving PET everywhere

The future of PET is bright—and the future is now. Affordable and upgradable-as-needed technologies ensure better access across facilities, reliable and reproducible exams that can be compared over time and reports that convey, in accurate numbers, the depth of disease and recommendations for treatment. Collectively, they are raising the value of PET.

Improving the technology is the first step to improving PET’s value. GE Healthcare debuted its DiscoveryTM IQ PET/CT system in the U.S. in September 2014. The system is designed to scan faster with lower radiation dose, provides excellent image quality and intelligent quantitation and was engineered to be economically accessible to virtually every PET center. Read that as scalable and upgradeable. The system delivers sensitivity up to 22 cps/kBq and a large axial field-of-view, up to 26 cm.

Specs are one thing, performance is another. Here’s where new thinking and innovation come in.

Imaging experts see the difference. “The technology makes it possible to perform exams with increased PET sensitivity while cutting the acquisition time and lowering the PET injected tracer dose” Courbon says.

“Physicians want reliable numbers, they want consistent SUVs, they want to use them,” Courbon notes. “Everybody knows that there are drawbacks with SUV and needs of improvements, but if you compared ‘basic’ SUV measurement of tumor response with the gold standard, which is survival, SUVs most of the time are strongly related to patient outcome. Even metrics that are not perfect are related to what is very important, the clinical outcome. We should be good with that. Imaging is working.”

Reporting Guidance for Oncologic 18F-FDG PET/CT Imaging

Clinical history

Indication for study; Cancer type and site, if applicable; Brief review of treatment history, if applicable.

Technique/procedure

Radiopharmaceutical name; Radiopharmaceutical dose/activity; Route of radiopharmaceutical administration; Uptake time (i.e., from radiopharmaceutical injection to imaging); Blood glucose level.

Ancillary medications administered, if applicable; Precise body region scanned; CT technique (including whether oral or intravenous contrast was used; if used, name and volume of agent).

Comparison studies

Whether comparison was made with prior PET or PET/CT studies; include dates when available; Whether correlation was made with prior non-PET imaging studies e.g., CT or MR imaging); include dates when available.

Findings

Location, size/extent, and intensity of sites of abnormal 18F-FDG uptake; Abnormal PET findings correlated with concurrent CT images or correlative; imaging studies, if applicable; Incidental PET findings and Incidental CT findings.

Impression

Clear identification of study as normal vs. abnormal; Interpretation of findings, rather than just restatement of findings; Succinct differential diagnosis provided, if applicable; Recommendations for follow-up studies, if applicable; Documentation of communication of urgent or emergent findings to referring physician or surrogate.

Source: J Nucl Med May 1, 2013 vol. 54 no. 5 756-761

Reference

¹Ahn S, et al., Quantitative comparison of OSEM and penalized likelihood image reconstruction using relative difference penalties for clinical PET.  Phys Med Biol 60:5733-5751, 2015.

²J Nucl Med May 1, 2013 vol. 54 no. 5 756-761

Disclaimer: This article was paid for by GE Healthcare, but none of the interviewees were paid for their participation in this article.