Profile: FDG-PET Whole Body
QIBA FDG-PET Whole Body
The Radiological Society of North America (RSNA) Quantitative Imaging Biomarker Alliance (QIBA) FDG-PET Technical Committee: Proffered Protocol for Quantifying the Standard Uptake (SUV) in Patients with Cancer
Version 1.0 of 19 June 2009
First draft composed by Andrew Buckler, Buckler Biomedical LLC, based on team inputs. Will be proffered by the FDG-PET Technical Committee.
- 1 Executive Summary, Introduction and Background Information
- 2 Claims (what users will be able to achieve)
- 3 Imaging Protocol: Overview
- 4 Subject Preparation
- 5 Imaging Procedures: General
- 6 Reconstruction and Reporting
- 7 Archival Requirements for Primary Source Imaging Data
- 8 Post-processing (i.e., anything not done on an acquisition platform that affects DICOM image data and/or pixel / voxel values)
- 9 Interpretation
- 10 Radiation Dose and Safety Considerations
- 11 Quality Control
- 12 Required Documentation
- 13 Acknowledgements
Executive Summary, Introduction and Background Information
The FDG-PET technical committee is composed of scientists representing the imaging device manufacturers, image analysis software developers, image analysis laboratories, biopharmaceutical industry, academia, government research organizations, professional societies, and regulatory agencies, among others. All work is classified as pre-competitive. A more detailed description of the FDG-PET group and its work can be found at the following web link: http://qibawiki.rsna.org/index.php?title=FDG-PET
The long-term goal of the FDG-PET committee is to qualify the quantification of anatomical structures with positron-emission tomography (PET) as biomarkers. The FDG-PET group selected metastatic cancer as its first case-in-point. The rationale for selecting metastatic cancer as a prototype is that the systems engineering analysis, the groundwork, profile claims documents, and roadmaps for biomarker qualification in this specific setting can serve as a general paradigm for eventually quantifying volumes in other structures and other diseases.
FDG is a glucose analogue. Its use in oncology is based on the fact that most types of tumors need more glucose than most other types of normal tissue (see Interpretation and pitfalls). F18-FDG is absorbed into the cell via glucose transporters in the cell membrane (Glut) and, in contrast to glucose, most of it is then trapped in the cell after phosphorylation by hexokinase. The kinetics of F18-FDG are such that static imaging produces adequate images approximately 60 minutes after administration. Not all tumors are F18-FDG avid: for example, many broncho-alveolar carcinomas have no Glut transporter upregulation. Furthermore, not all patients with a particular type of tumor will have an F18-FDG-positive tumor: this applies to e.g. renal cell cancer, thyroid cancer and carcinoid cancer. The reason for this variability is not entirely clear.
PET is an 'evolving' field at both a national and international level, with sometimes striking differences between individual countries. The summary below is therefore subjective in nature and based on a combination of experience and literature (1-13).
- primary presentation
- diagnosis: unknown primary malignancy, solitary pulmonary nodule (in the case of a discrepancy between the clinical and radiological estimates of the likelihood of cancer (2));
- staging on presentation: non-small-cell lung cancer (3), T3 esophageal cancer (4), Hodgkin's disease (generally when it is unclear whether the disease has extended beyond the neck/armpits/mediastinum), locally advanced cervical cancer and perhaps also locally advanced breast cancer / ENT tumors with risk factors.
- response evaluation: malignant lymphoma, GIST, at present other applications only in a research setting. Esophageal cancer, lung cancer and breast cancer appear promising.
- restaging in the event of (suspected) relapse (for tumors that are likely to accumulate F18-FDG)
- elevated serum markers (e.g. colorectal, thyroid, ovarian, cervix, melanoma, breast and germ-cell tumors)
where local treatment has been carried out to treat a limited relapse that may be associated with high morbidity (e.g. liver/lung metastases
This protocol should be considered for use in the care of individual patients in conventional medical settings, as well as in clinical trials of new therapies for cancer. An example of these clinical settings are described in Table 1 for lung cancer. Separate Profiles describe processes and procedures for quantifying SUV in different clinical settings.
Relationship with other diagnostic methods
Deviations on a PET scan must be correlated with the CT scan (MRI where appropriate). These kinds of test will in many cases already be available (and of course this is always the case with PET-CT scans). The aim of this correlation is to find an anatomical substrate (especially of the clinically significant PET findings), mainly in order to minimize false-positive results. In a direct sense this is the case where the CT scan provides a clear benign explanation for the FDG-positive focus. But if the corresponding deviation is suspect on the CT, the referring physician can be given clear information as to the biopsy site. If no substrate can be found on an adequate CT or MRI (technically adequate, suitable for the area being assessed and carried out not too long ago), then that deviation may be regarded as a false-positive. This ensures that patients presenting at the primary stage are never unjustly deprived of treatment with a curative aim.
Covariate requirements for accurate and precise SUV quantification
- General proposal is to
- Improve and standardize measurements affecting SUV quantification (e.g. time, weight)
- Embed results in public DICOM tags of patient study
- Display results on viewing and analysis workstations (glucose, FDG uptake period)
- Patient weight:
- Rationale: Patient weight is required for the standard SUV body weight correction and should be measured on the day of the PET/CT scan. Entry of weight is also required for phantom studies to verify accurate scanner calibration and SUV calculation.
- Limitations: The precision that can be recorded on some PET/CT scanners is only 1 pound or kilogram. This leads to errors for small patients (e.g. pediatric) and more importantly phantom measurements. Patient weight is sometimes automatically imported from medical records and is not necessarily accurate.
- Proposed Solution: All PET/CT vendors should provide an override of automated weight import and capability to allow entry of weight with a minimum precision of +/-0.1 pounds or kilograms. Patient weight should be measured on a properly calibrated scale in the imaging department on the day of each scan.
- Patient height:
- Rationale: Accurate patient height is required for SUV body habitus corrections such as body surface area (BSA) and lean body mass (LBM).
- Limitations: Height is often an optional measurement and can have different values at different visits.
- Proposed Solution: Height should be a mandatory measurement required at the time of acquisition. Software used for longitudinal analysis should produce warning notification for inconsistent height values in the same patient and force the use of a single value.
- Injected activity:
- Rationale: Accurate recording of the net injected activity is essential for measuring accurate tumor SUVs
- Limitations: The acquisition software for some PET/CT scanners does not allow the entry of residual activity and calculation of net injected activity. Many customer sites do not measure residual activity and/or properly correct for decay. The assay dose and/or time are often incorrectly used as net injected activity and injection time.
- Proposed Solutions: All PET/CT scanners should record assay activity and measurement time; residual activity and measurement time; and time of injection. The PET/CT scanners should automatically calculate the net decay-corrected injected activity. The assay and residual activities and times should be stored in public DICOM tags (in addition to the net injected activity that is currently stored) in association with each patient PET/CT scan.
- Blood Glucose:
- Rationale: Blood glucose can affect the biodistribution of FDG and the quantification of tumor SUVs.
- Limitations: Most PET/CT acquisition software does not record blood glucose and there is no public tag for the data.
- Proposed solutions: PET/CT acquisition software should record the blood glucose value and time of measurement in a public DICOM tag.
- Rationale: Accurate time is essential to ensure proper decay correction and accurate SUV quantification
- Limitations: PET/CT scanners often have multiple computers that are not synchronized with each other and/or absolute time standards (e.g. NIST). Power disruptions and computer failures can result in loss of time. Daylight savings time changes that are not properly accounted for can result in one hour errors that translate into 32% SUV errors.
- Proposed solutions: PET/CT scanner vendors should incorporate daily time calibrations with absolute standard on each computer. Customer sites should be educated on the importance of calibrating all clocks and watches including those used in the hot lab used to measure activity. For F-18, +/- 1 minute is sufficient but isotopes with shorter half-life will require +/5 seconds.
- Fasting time
- Rationale: Insufficient fasting time can affect FDG biodistribution and SUV quantification.
- Limitations: There are no mechanisms for recording or reviewing fasting time in DICOM images.
- Proposed solutions: Fasting time and state should be recorded during PET/CT acquisition and recorded in public DICOM tag associated with each patient PET/CT study.
- Rationale: Insufficient hydration can affect FDG biodistribution and potentially SUV quantification.
- Limitations: There are no mechanisms for recording or reviewing hydration in DICOM images.
- Proposed solutions: Hydration status (number of 12 oz glasses of water consumed in the past 12 hours) should be recorded during PET/CT acquisition and recorded in public DICOM tag associated with each patient PET/CT study.
- Physical Activity
- Rationale: Excessive or strenuous physical can affect FDG biodistribution and potentially SUV quantification.
- Limitations: There are no mechanisms for recording or reviewing physical activity assessments in DICOM images.
- Proposed solutions: Physical activity status should be recorded during PET/CT acquisition and recorded in public DICOM tag associated with each patient PET/CT study.
- FDG Uptake Time
- Rationale: Different FDG uptake times particularly in serial assessments of tumor metabolism can significantly affect SUV quantification. This includes patient orientation for acquisition.
- Limitations: There are limited mechanisms for recording or reviewing FDG uptake times on display and analysis workstations.
- Proposed solutions: Calculated FDG uptake based on injection time and initiation of scanning time should be available for review on the display and analysis workstation. In addition the patient orientation during acquisition should be displayed.
Claims (what users will be able to achieve)
Quantitation, both single time-point and longitudinal*, of tumor metabolism via FDG-PET/CT that can be used practically, reliably, and efficiently as a biomarker. This biomarker shall be of known precision and accuracy when used in single- or multi-center, multi-vendor, clinical trials and meta-analyses, with enhanced potential for use in patient care and qualified for use in clinical studies of pathophysiology and therapeutic interventions.
- i.e., measuring change over time, e.g. pre- and post-treatment in a clinical trial of an investigational agent or for refinement of patient care during a course of therapy.
[NB; This may in future include meta-analysis across trials and databases, and may be more broadly applicable to other “final common pathway” tracers of the hallmarks of cancer e.g. proliferation, apoptosis, angiogenesis, any of which would improve confidence around expanding the indications of PET/CT in response monitoring for both clinical trials and patient care.] [NB; Quantification includes proper calibration and resolution (partial volume effect) but will be tackled later along with other issues from the original brainstorming session. Specification of performance targets might be based on pharma needs to reduce the number of patients required in clinical trials for a given effect size. In order to do this, it is necessary for vendors to measure and report performance metrics for SUV.]
- 1st; measure (accurate and precise)
- 2nd; change metric (longitudinal)
- 3rd: response metric (clinical relevance)
Claim #1: Can Can create, store, retrieve FDG-PET/CT images (? and attenuation scans;? specify organ/tumor type)
- Precursor: Any change on Image generation SW (reconstruction, normalization, calibration, correction methods) and any manipulation SW (registration, ROI definition, or SUV calculation, calcium scoring) will impact the final quantitation. This applies to both PET and CT subsystems. Typically, following are the SW applied to PET/CT domain. Some manufactures may combine some of those SW into one tagged version.
- Acquisition/console SW (may include Reconstruction)
- Reconstruction SW (may include off-line recon and multiple methods)
- Workstation SW (may have the same version number as display/quantitation SW)
- Display/quantitation SW (may be as option CD or third party SW)
Due to the continuous improvements of current PET/CT systems, vendors will upgrade above types of software for both PET/CT subsystems from time to time in any part of sub-systems or systems. Although not every SW upgrade will impact the quantitation, some of upgrades will include quantitation improvements. Hence, it is very important to inform the user regarding the update version and its impact.
- Stage I: Manually record the SW version number and flag the impact to quantitation – Time: Immediately
Currently, only acquisition (image generation) SW can be stored in DICOM header. Other SW version numbers can be obtained manually. The instructions are usually in the user’s manual or service engineer’s manual. Trial team responsibility: Create a recording sheet (or mechanism) for technologists to record any SW version (change) after upgrades. This information will passed to QC team for assessment and proper follow-up action (This is a gap QC team need to address) Manufacture responsibility: 1) Provide step-by-step instructions for obtaining the SW version numbers for all sub-systems, 2) When any update is performed, inform the user on the impact on quantitation by simply indicating “Yes or No”. 3) Publish DICOM conferment statement and provide technical support to trial team if it decides to obtain the version number by using SW. QIBA Responsibility: Inform the manufacture the need and ask for corporation
- Stage II: Record all SW Version numbers in DICOM header
By requesting additional DICOM tags, a QC tool could automatically check the status of SW version and its impact to quantitation – Time: 3-5 year (1-3 for DICOM committee to add tags and 2-3 year for manufactures to comply). This information could be modeled into trial data. Trial Team Responsibility: 1) Create a SW which can obtain and summarize the version numbers and quantitation information; 2) incorporate the information into QC model (QA team gap??) statistically. Manufacturer Responsibility: 1) Record all SW versions into DICOM tags, 2) Record quantitation impact tag; 3) provide variance on SUV. QIBA Responsibility: Define how many additional DICOM tags are needed and request DICOM tags to DICOM committee.
Claim #2: Can reproducibly create, store, retrieve SUV measurements (? specify organ/tumor type)
- Precursor: (granularity from DRO Team)
- Rationale: (granularity from DRO Team)
Claim #3: Can create, store, and retrieve mark ups of SUV measurements in accordance with a standard definition for RoI that provides a known balance between precision and accuracy
- Precursor: A survey has been prepared for workstation vendors to allow assessment of the ROI tools and algorithms currently available. Topics include the specifics of how individual pixels are assigned to ROI’s, how multi-slice ROI’s are defined, which statistics are reported, and how the resulting numbers can be stored or transferred from the system. The results of this survey will be used to determine the degree to which common methodologies can be defined on current systems, and what recommendations to make to vendors for common ROI methodology in the future.
- Rationale: All PET image display and analysis workstations, including those provided by the scanner manufacturers and those from third parties, include Region of Interest (ROI) analysis capability. The statistics returned from the ROI analysis form the basis of PET quantification. The current widespread standard is for the user to identify a region that surrounds the highest-uptake area of a tumor with a geometric (generally circular or spherical) ROI, and record the maximum pixel value within this region. This method allows two operators to derive the same value as long as the maximum pixel is somewhere within the drawn ROI. Potential drawbacks included the higher variance that may come from using only a single pixel value. Clinical trials and routine clinical practice both need well-defined ROI techniques that work similarly across workstation platforms.
- Work plan:
- A variety of ROI methods have been proposed that would, in principle, improve the variance of the max-pixel method by additionally reporting the mean of a pre-defined cluster of neighboring pixels (e.g., all pixels within a 1 cm circular or spherical region.) Such methods are in wide-spread enough use that it would be appropriate to recommend their adoption by workstation vendors.
- For a given cluster of pixels used for calculated a mean value, it would be equivalent to smoothing the image with a kernel based on this cluster shape and size, and then simply report the maximum pixel value within a broad, user-defined region. The technical and implementational benefits of using a ROI mean versus the peak pixel from a smoothed image should be investigated further, specifically the precision and accuracy trade-offs.
- PET/CT allows, in principle, anatomically based ROI’s that could be applied to PET. Such methods could provide objective means of measuring FDG uptake within larger volumes. Potential issues include that actual determination of the lesion on the CT images, misregistration between the PET and CT, and interpolation that would be required to translate from CT image space to PET image space. The issues of CT-based ROI’s need to be investigated, including issues of misalignment between PET and CT, potential corrections, and interpolation schemes. In addition, the potential for correction of the effects of diminished ROI values due to limited PET spatial resolution, based on CT information, as has already been proposed, should be further investigated.
Claim #4: Can find in the DICOM header reliable data required for SUV calculation
- Precursor: vendors have supplied DICOM header info and conformance statements(untrusted)samples (test/retest) actual images of NEMA phantom from multiple vendors and systems understanding of physics to produce synthetic phantom experience with use of DROs for reconstruction scheme evaluation
- is no existing DRO for QC of sites/systems
- is no way to readily compare SUV from different analysis software
- physical phantoms are not sufficient to solve this class of problem, even though they obviate the need to continually update a DRO for each new product version
- Solution: DRO group to merge with SUV calculation group, because cannot validate calculation based on header without pixel data source and known expected values. Goals:
- Short term (threshold) goal – single standard stack (or one for each vendor based on what we know).
- Middle term (target) goal – validate receiving workstations with vendor-specific stack with header (and standardized SUV truth values) – supply floating point data, abstract meta-data, vendor produces pixel data and DICOM header and series organization – if vendor can provide multiple forms (e.g., in SUVbw or MBq/ml or PROPCNTS) provide all of them (overlap with s/w version group – needs to tell us what changes and how it affects computation) – supply pre-defined ROIs with pre-defined truth values for them – need to define supplied slice thickness, FOV and resolution, including cross-plane for 3D SUV calculation (i.e., on coronal or sagittal recons or using a 3D (cylindrical or spherical)) – CT data NOT needed unless want to validate regional measurements as opposed to point measurements (a refinement to the middle term (target) goal) - want two phantoms – one with no noise so that the values are absolutely predictable, and with noise for realism (want asymmetric data for regional calculation check). NOT A PRODUCT FEATURE BUT SAMPLE DATA ONLY.
- Long term (ideal) goal – validate image generation beyond acquisition point AND receiving workstation - reconstruct from standard raw data (with provided reconstructed attenuation (not diagnostic) CT image), prior to any corrections (sinugram, list mode) - measures software performance ONLY, not hardware problems (non-goal – need real phantoms to detect such changes); CT is required – need to define corresponding slice thickness, FOV and resolution (non-goal is testing registration fusion, which requires a different type)
Use NEMA NU2 phantom as basis for synthetic phantom data.
- Additional use-cases:
- feeding other vendors consoles with their own phantom data (e.g., from different versions of their own hardware)? What about other vendors’ data on a vendor’s console ?
- what about phantoms to validate registration, and the impact on SUV, especially deformable registration
Include software version info? Include calibration info ?
Claim #5: Delta of 10% to discern dose effect in phase II trials, delta of 25% to discern clinically relevant change in phase III trials. The specific aim of this image acquisition and processing protocol is to describe procedures that seem sufficient for quantifying the SUV in conduct of multi-center trials. Answers to questions such as how hard is it to acquire, quantify, aggregate, analyze, and interpret the data? Logistics, user-friendliness, number of clicks to get data, likelihood of technical achievement consistently across sites, and the efficiency in conduct of multi-center trials. Issues seen as critical to quality (CTQ) for pharmaceutical development include how big a trial is required to demonstrate performance of our drug (phase III). Smaller sigma enables smaller clinical trials. Two points on a line will enhance likelihood of comparability across a dynamic range relevant to both the evaluation of “response” across populations in clinical trials and the individualization of therapy on a per-patient basis. CTQs for Regulatory include clinical relevance of results from multi-center trials; combinability of data across sites (do the results from each center mean the same thing), is heterogeneity across sites so great that this is not a single trial but a meta-analysis?, interpretability of data for drug approval/labeling, and can the regulatory reviewer reasonably assume these data represent beneficial treatment effect which support the use of this drug in a well-characterized patient population?
- Precursor: Demonstrate this accuracy and repeatability is achievable
- Groundwork: Test-Retest measurements of phantoms, i.e. very-best-case-scenario, with variability one order of magnitude less than variability in "real-life", i.e. algorithm returns variability of less than X%
<Relevant Groundwork Link 2:> Test-Retest measurements of small sample of NIST cases, i.e. nearly-best-case-clinical-scenario, with variability for measurement of isolated, simple tumors of less than 2X% (up to 4 times the noise in phantoms and less than one fifth the noise expected in real life scenarios).
<Relevant Groundwork Link 3:> Test-Retest measurements of a few well behaved masses in coffee break studies of less than 4X% between Image Set 1 and Image Set 2 of each patient studied twice in succession. This 10% threshold is somewhat capriciously based on the assumption that the precision of measurement in selected coffee break tumors will be twice as good as that which can be achieved in most clinical trial scenarios.
Imaging Protocol: Overview
The protocol describes, in predominantly chronological order, procedures that are required to achieve this level of precision. At each step in the process, key terms include procedures that are (1) "acceptable", by which it is meant that anything less rigorous will likely fail to meet minimum criteria for precision and accuracy; (2) "targets", by which it is meant parameters that are most likely to meet or exceed specifications; and (3) "ideal", by which it is meant parameters that are probably the best fit-for-purpose currently available, regardless of the effort required to implement them.
The protocol describes procedures that should be universally followed in this setting, regardless of the instrument that is used to acquire the data. It also provides links to tables that list specific settings on various makes-and-models of PET scanners.
Utilities and Endpoints of the Imaging Protocol within the Clinical Trial
- This image acquisition and processing protocol should be sufficient to quantify the SUV of a malignant tumor, and its longitudinal changes within subjects. The protocol is otherwise agnostic to the clinical settings in which the measurements are made and the way the measurements will be used to make decisions about individual patients with cancer or new treatments for patients with cancer.
Management of Pre-enrollment Imaging Tests
- The history of prior medical imaging procedures that might, or might not, be used as part of the selection criteria for enrolling patients in a clinical trial that uses this protocol is outside the scope of the QIBA FDG-PET committee. However, only image acquisition and processing protocols that conform to, or exceed, the minimum design specifications described in this protocol are sufficient for quantifying SUV with the precision of measurement specified in the Claims section. In practice, this will often require "baseline" scans to be repeated according to these guidelines when the objective is to quantify longitudinal changes within subjects.
Timing of Imaging Tests within the Clinical Trial Calendar
- The FDG-PET committee agrees with the authors of RECIST 1.1., who wrote " all baseline evaluations should be performed as close as possible to the treatment start". Otherwise, these imaging procedures are not time-sensitive. The interval between follow up scans within patients may be determined by current standards for good clinical practice (cGCP) or the rationale driving a clinical trial of a new treatment.
Management of On-protocol Imaging Performed Off-schedule
- "On Protocol, Off Schedule" PET scans should be acquired, processed, and analyzed exactly like on protocol, on schedule PET scans.
Management of Off-protocol Imaging
- The FDG-PET committee notes that other sources of information, including Off-Protocol imaging procedures, can add valuable information about the management of individual patients or the conduct of a clinical trial. However, their use is outside the scope of this image acquisition and processing protocol.
Subject Selection Criteria Related to Imaging (mainly exclusionary in nature)
- There are few, if any, absolute contra-indications to the PET-CT image acquisition and processing procedures described in this protocol.
- The FDG-PET committee recognizes that there may be relative contra-indications, e.g., in young children or pregnant women. Methods for quantifying and classifying relative risks are referenced. Otherwise, explications of risks are predominantly outside the scope of work conducted by the FDG-PET technical committee.
- The FDG-PET committee acknowledges that there are potential risks associated with the use of contrast material. The default recommendations for intravenous and oral contrast that follow assume there are no known contra-indications in a particular patient. The committee assumes that local standards for good clinical practice (cGCP) will be substituted for the default in cases where there are known risks, e.g., patients with chronic renal failure.
Detail: Interval Timing (e.g., oral and/or IV intake, vigorous physical activity, timing relative to non-protocol-related medical interventions, etc.)
- There are no specific patient preparation procedures for the CT scans of the chest described in this protocol. The FDG-PET committee acknowledges that there are specifications for other procedures that might be acquired contemporaneously, such as requirements for fasting prior to FDG PET scans or the administration of oral contrast for abdominal CT. Those timing procedures may be followed as indicated without adverse impact on these guidelines.
Detail: Specific Pre-imaging Instructions
Detail: Prior to Arrival
- The local standard of care for acquiring PET-CT scans may be followed. For example, patients may be advised to wear comfortable clothing, leave jewelry at home, etc.
Patients are not allowed to consume any food or sugar for at least four hours prior to the start of the PET study. In practice, this means that patients scheduled to undergo the PET study in the morning should not eat after midnight and those scheduled for an afternoon PET study may have a light breakfast before 8.00 a.m. Medication can be taken as normal. Adequate pre-hydration (for example, one litre (!) of water in the two hours prior to injection; where necessary, take the volume of water in oral contrast medium for a diagnostic CT scan into account) is important to ensure a sufficiently low urine F18-FDG concentration (less chance of artefacts) and for radiation safety reasons. The infusion used to administer intravenous pre-hydration must not contain any glucose. Patients must avoid extreme exertion before the PET study (for example, they must not cycle to the hospital). The following rules apply to patients with diabetes mellitus:
- type II diabetes mellitus (controlled by drugs)
- the PET study should preferably be performed in the late morning
- patients must comply with the fasting rules indicated above
- patients may continue to take oral substances to lower their blood sugar.
- type I diabetes mellitus and insulin-dependent type II diabetes mellitus
- ideally, an attempt should be made to achieve normal glycaemic values prior to the PET study, in consultation with the patient and his/her attending MD
- the PET study should be scheduled for late morning
- the patient should eat a normal breakfast at 7.00 a.m. and inject the normal amount of insulin. Thereafter the patient should not consume any more food or fluids, apart from the prescribed amount of water.
In the case of patients on continuous insulin infusion, the PET study should if possible be scheduled early in the morning. The insulin pump is kept on the "night setting" until after the PET study. The patient can have breakfast after the PET study. A bladder catheter is placed only if required, and this should preferably be done before F18-FDG is administered. Administration of a diuretic (furosemide) can be considered in the case of small pelvic tumors, but this should not be routine practice.
There is no reason for routine administration of sedatives (diazepam, temazepam). Sedatives can be considered in the case of tumors in the head and neck region to reduce muscle uptake. These should then be given about 60 minutes before administration of F18-FDG. Benzodiazepines are not helpful in avoiding or reducing the so-called brown fat phenomenon.
Blood glucose level must be measured prior to administering F18-FDG. A Glucometer® or a similar device (capable of performing overall euglycaemia measurements) can be used for this purpose, but a blood glucose test must be performed in the hospital's own laboratory using a calibrated and validated method in the case plasma glucose level is used as correction of SUV measurements.
- if glucose is < 11 mmol/l the FDG PET study can be performed
- if glucose is ≥11 mmol/l the FDG PET study must be rescheduled
N.B.: insulin must not be given to reduce glucose levels (this leads to greater muscle uptake of FDG) unless the interval between administration of insulin and administration of FDG is more than four hours. The ambient temperature in the administration room / the room in which the patient is left to recover must be comfortable (it may be necessary to provide extra blankets) in order to avoid uptake in brown fat (see 11. Interpretation and pitfalls).
Detail: Upon Arrival (including ancillary testing associated with the imaging and downstream actions relative to such testing)
- Staff shall prepare the patient according to the local standard of care.
- Patients should be assessed for any removable metal objects on their bodily surfaces that will be in the field of view.
- Patient should be "comfortably positioned", in "comfortable clothes to minimize patient motion and stress (which might affect the imaging results) and any unnecessary patient discomfort.
- Bladder State
- Ideal: micturation immediately prior to being placed in the gantry
- Target: empty bladder
- Acceptable: any
- The target here is purely for patient comfort.
- Note: Factors that adversely influence patient positioning or limit their ability to cooperate should be recorded in the corresponding DICOM tags and case report forms, e.g., agitation in patients with decreased levels of consciousness, patients with chronic pain syndromes that limit their ability to cooperate with requirements for breath holding or remaining motionless, etc.
Imaging Procedures: General
Detail: Data that should accompany the request for a PET studya. Indication, reason for request of PET study (see 2. Indications)
- Height and weight (these must be determined precisely in the case of SUV measurements, see below)
- (If known) tumor type, tumor sites that have already been noted
- (Oncological) prior history
- Diabetes mellitus (medication)
- Results of other imaging tests (especially CT/MRI).
Detail: Imaging Agent Preparation and Specification (Contrast agent or radiopharmaceutical) The FDG-PET committee acknowledges that the use of intravenous contrast material is often medically indicated for the diagnosis and staging of lung cancer in many clinical settings. However, the use of contrast is not an absolute requirement for quantifying the volumes of many tumors in the chest or achieving the specific aims associated with this protocol.
- Product : F-18-fluorodeoxyglucose (F18-FDG)
- Nuclide : Fluoro-18
- Dosage : Dependent on the scanner used and the patient's weight. (See 7. Performance).
- Administration : Intravenous
Detail: Contrast administration: (agent, dose, route) Essential data/aspects and required materials for the FDG PET study
- Patient's height and weight (to be measured for SUV).
- When did the patient last receive chemotherapy, G(m)-CSF and/or other growth factors? The protocol requires the minimum interval between administration of such substances and the PET study to be defined. In the case of patients undergoing regular treatment, the minimum period between the last dose and the PET study should be 10 days if possible.
- (Ideally) a triple-channel system for administering the tracer and flushing with physiological saline; bedside glucose meter (for checking overall serum glucose especially in patients susceptible to hyperglycaemia (diabetics, patients taking corticosteroids)). N.B.: these bedside methods are not suitable for normalising/correcting SUV.
Detail: Procedure for performing the PET study
- An indwelling intravenous device (e.g. Venflon) is used to administer the F18-FDG intravenously once the patient's blood glucose has been determined and blood samples for laboratory testing have been taken if necessary. Make sure that if there is a needle on the syringe it is free from FDG.
- Flush and rinse out the administration syringe with at least 10 cc of normal saline using the three-way valve.
- The venflon can be removed after intravenous administration (unless CT contrast agent is to be administered subsequently by intravenous injection).
- The waiting room must be relaxed and warm. Give the patient extra blankets if necessary.
- Tell patients to lie or sit as calmly as they can, and not to talk unnecessarily. They may go to the toilet while waiting.
- During the waiting period patients will be asked to drink another half a litre of water, or this amount can be given in the form of physiological saline administered by infusion or catheter.
- Ask the patient to urinate a few minutes before the start of the PET study.
- The interval between FDG administration and the start of acquisition is 60 minutes. When repeating a scan on the same patient, try to apply the same interval (tolerance ±5 minutes).
- A 'whole-body' uptake normally covers the part of the body from the mid-femora to the external auditory meatus (in that direction, as bladder activity increases during the scan). A longer scanning trajectory may be used if appropriate.
- Scan acquisition depends on various factors, including the scanner type (PET, PET-CT) and the acquisition mode (2D, 3D). For CT settings in the case of PET-CT, CT whole body or low-dose CT, see 9. Camera and computer. Follow the supplier's recommendations (see table 2 for an example). Transmission scanning time for each bed position: depends on whether the scan is a CT scan or a transmission scan with Ge-68/Ga-68 source.
- In the case of PET-CT, the patient is usually in supine position with the arms raised.
In general, PET-CT is carried out using a protocol comprising a scanogram/scout scan/topogram and a low-dose CT for attenuation correction (CT-AC) and anatomical correlation. IV contrast agent must not be administered during the low-dose CT because of its influence on SUV calculation.
- In the case of dual-slice CT, artefacts are created in the diaphragm area when the patient breathes. The patient must therefore hold his/her breath for a few seconds on the technician's instructions during CT-AC acquisitions. No such instructions need be given in the case of PET-CT scanners with more than two slices. The CT-AC scan can then be carried out while the patient continues to breathe shallowly.
- There is as yet no generally accepted oral contrast method for PET-CT scans. In the case of SUV measurements of intra-abdominal lesions it should be borne in mind that (oral) contrast agents affect the SUV outcome (10-12): intestinal preparation (with diluted 'positive' oral intestinal contrast agent) can produce quantitative inaccuracy of up to 20% (10, 11), and is therefore not recommended at present. Artefacts that can impair visual assessment have also been described (12). There is as yet insufficient experience with the use of negative oral contrast agents (13). These substances are therefore not (yet) the preferred choice.
- A standard diagnostic CT scan with (i.v.) contrast agent may if appropriate be carried out according to standard radiological methods after the low-dose CT and PET acquisition.
- F18-FDG dosage assuming a fixed scan duration of 5 min per bed position and a bed overlap of less than 25%:
In the case of 2D scans: ca. 5 MBq/kg body weight (± 10 %). In the case of 3D scans: ca. 2.5 MBq/kg body weight (± 10 %).
- In section 8 permitted alterations of the protocol are given.
- Specifications of transmission scans based on a Ge-68 line source: > 3 minutes per bed position. Practitioners are advised to use the scanner manufacturer's settings for CT-based attenuation correction.
Detail: Protocol alterations permitted in the case of multi-centre studies
- When using scanners with a high count rate capability (LSO, LYSO and GSO-based cameras with or without ToF), the dosage and scan duration for each bed position must be adjusted so that the product of the dose and scan duration +10% (see 8.5) is equal to or greater than the specifications set out below. Therefore, one may decide to apply a higher dosage and reduce the duration of the scan.
- The figures for scanners with bed overlap of < 25% (Siemens and GE) are:
- Product of MBq/kg x min/bed > 27.5 for 2D scans
- Product of MBq/kg x min/bed > 13.8 for 3D scans The dosage is then calculated as follows:
- Dose for 2D scans = 27.5 x weight / (min/bed)
- Dose for 3D scans = 13.8 x weight / (min/bed)
- Minimum scan duration = 3 minutes
- And for scanners with a bed overlap of 50% (Philips):
- Product of MBq/kg x min/bed > 6.9 (3D only)
- Dose = 6.9 x weight / (min/bed)
- Minimum scan duration = 2 minutes
- The dose may be increased (to save on scanning time) only if the higher dose leads to count rates that are well within the scanner's count rate capacity. Dose elevation must not cause quantification problems. This is a matter for which each centre is individually responsible. They must demonstrate that the dose and scanning times are acceptable by means of phantom tests or published data relating to the scanner in question (where appropriate, on the basis of scanner specifications measured by the participating institution).
- It must be shown that the use of a higher dose (combined with a shorter scanning time) does not impair image quality as a higher dose usually leads to lower 'noise equivalent count rates' (NECR). This aspect is also covered by the requirement to increase the product of dosage and scanning time by 10%. A higher dose causes a disproportionate increase in the randoms contribution, which in turn leads to a lower NECR and therefore potentially poorer image quality.
- In the case of BGO scanners, the specified dose as indicated under 7p must not be increased.
- If the scanning duration for each bed position can be set separately, then the scanning duration per bed position may be further reduced by up to 50% for bed positions outside the thorax and abdomen (i.e. at the level of the head, neck and legs, as attenuation is less). The dose must still be calculated assuming the scanning duration per bed position as used for bed positions at the level of the thorax and the abdomen.
The specifications indicate that heavier patients receive a higher dose. A short scanning duration per bed position should also be offset by a higher dose. Two model calculations are given below to clarify the situation.
Calculation of dose to be administered, example 1:
- 3D scanner with a bed overlap of less than 25% (e.g. Siemens and GE cameras).
- Patient weighing 70 kg
- Scanning duration per bed position: 3 minutes per bed The dose to be administered is therefore:
13.8 (MBq/kg per min/bed) x 70 (kg) / 3 (min/bed) = 13.8 x 70 /3 = 322 MBq.
Calculation of dose to be administered, example 2:
- 3D scanner with 50% bed overlap (e.g. Philips cameras).
- Patient weighing 70 kg
- Scanning duration per bed position: 3 minutes per bed The dose to be administered is therefore:
6.9 (MBq/kg per min/bed) x 70 (kg) / 3 (min/bed) = 6.9 x 70 /3 = 161 MBq.
Table 1 presents the specifications set out in section 7 and the section above. The dose to be administered is given in MBq.
Table 1. Mode 2D 3D 3D 3D 3D 3D 3D 3D 3D Time/bed 5 5 4 3 2 5 4 3 2 Bed overlap < 25% < 25% < 25% < 25% < 25% 50% 50% 50% 50% Scanner HR+,2D HR+,3D HR+,3D HR+,3D HR+,3D Gemini Gemini Gemini Gemini Discovery, 2D Discovery Biograph Biograph Biograph GeminiToF GeminiToF GeminiToF GeminiT Patient weight (kg) Biograph 30-39 170 80 120 160 240 40 60 80 120 40-49 220 110 150 200 310 60 70 100 150 50-59 270 130 180 250 370 70 90 120 180 60-69 320 160 220 290 440 80 110 140 220 70-79 370 180 250 340 510 100 120 170 250 80-89 420 210 290 390 580 110 140 190 290 90-99 470 230 320 430 650 130 160 210 320 100-109 520 260 360 480 720 140 180 240 360 110-119 570 280 390 520 790 150 190 260 390 120-129 620 310 430 570 860 170 210 280 430 130-139 670 330 460 620 930 180 230 310 460 140-150 720 360 500 660 1000 200 250 330 500 N.B.: the scanners mentioned are given as examples. The key factors in determining the dose are the mode of acquisition (2D or 3D), the time per bed position (time/bed) and the degree of bed overlap.
- Ideal: (a) Contrast should be administered in a dynamic fashion, preferably with a power injector. At baseline and at each subsequent time-point, the same dose of contrast and rate of contrast administration should be performed as clinically safe. (b) Scan delay after contrast administration is dependent upon the both the dose and rate of administration, as well as the type of scanner being used. Generally institutional guidelines should be followed so as to optimize reproducibility of the scan technique.
- Target: Same rate and dose of contrast administration, and exact same start time of scans relative to contrast administration. Sites should use the same brand of contrast each time they scan a particular patient.
- Acceptable: Manual administration or no administration. Regardless, exactly the same contrast agents and administration procedures must be used in each examination, even if that means no intravenous contrast is ever given.
Contrast Dose Reduction Based On Creatinine Clearance: (renal function) Site-specific sliding scales that have been approved by local medical staffs and regulatory authorities should be used for patients with impaired renal function.
Detail: Imaging Data Acquisition
- Camera and computer
- Other acquisition parameters
- Emission scans must be conducted using the following additional settings:
- online randoms correction based on ‘delayed coincidence time window’ technique
- indication of the correct isotope, the patient's height and weight, and the dose administered
- decay correction must be 'on' (see any reconstruction settings) to allow for working back to the start of the scanning time, T = 0.
A 'dedicated' CT scan for attenuation (the 'CT-AC') must be carried out in accordance with the scanner manufacturer's recommendations (default settings).. The following additional provisions also apply:
- no contrast agent may be used before or during the CT-AC scan as it will cause artefacts in the PET scans. If a diagnostic (i.v.) contrast-enhanced CT is performed to complete the procedure, this must be done after the CT-AC and the PET studies have been finished.
- ensure that the patient is lying within the CT-AC field of view (FOV).
- In some PET-CT scanners, the FOV of the CT and CT-AC is smaller than that of the PET. Truncating the CT (and CT-AC) causes reconstruction artefacts and therefore inaccurate quantification of the PET scan.
- When using Ge-68 transmission sources, they must be replaced on time (i.e.: at least once every eighteen months).
- The dosage must never exceed the maximum dose as recommended by the supplier. If this is not specified, then the dose applied must not lead to the camera's maximum count rate being exceeded (dead time).
Table 2. Example of CT acquisition settings for a Siemens Biograph (see manufacturer's instructions for each scanner)
Tube position AP
Length (mm) 1024
Scan direction Cranial -> Caudal Eff. mAs 40 (> 100 kg: 50, > 120 kg: 60)
Care dose (on/off) Off
Rotation time (sec) 0.8
Slice width (mm) 6.0
Slice collimation (mm) 5.0
Feed/Rotation (mm) 15.0
Scan direction Caudal -> Cranial
Delay (s) 3
Breathing instructions Hold breath when instructed (only for dual-slice-CT)
Arms Raised FOV (mm) 500
Slice width (mm) 6.0
Recon increment (mm) 3.0
Image order Cranial -> Caudal
Window 400/40 abdomen
Kernel B30s medium smooth
Detail: Subject Positioning The following details shall be recorded, manually by the staff if necessary: Patient Positioning
- Ideal: Patients should be positioned in the exact same way for every scan, with careful attention paid to details such as the position of their upper extremities, the anterior-to-posterior curvature of their spines as determined by pillows under their backs or knees, and the lateral straightness of their spines. For patients placed in a prone position, the head should be tilted in the same direction each time.
- Target: Supine/Arms Up/Head First
• Note: Target is provided as a default to drive some consistency when details of prior scans are not available.
- Acceptable: any position that is consistent with prior scan, e.g., in patients who are physically unable to have their arms placed above their heads, every effort should be made to insure that the upper extremities lie on the table in the same position each time.
• Consistency is required to avoid unnecessary variance in attenuation, changes in gravity induced shape, or changes in anatomical shape due to posture, contortion, etc. When changes in position are medically unavoidable due to a change in clinical status, details should be provided.
Detail: Instructions to Subject During Acquisition (e.g., breathing) Breath Hold
- Ideal: Single breath hold at full inspiration
- Target: Single breath hold at full inspiration
- Acceptable: suspended respiration near high % of end inspiration
• Breath hold reduces motion, which degrades the image.
• Full inspiration inflates lungs which is necessary to separate structures and make lesion more conspicuous.
Detail: Timing (e.g., relative to previously administered imaging agents / enhancers; inter-time point standardization)
The time-interval between the administration of intravenous contrast and the start of the image acquisition should be determined in advance, and then maintained as precisely as possible during all subsequent examinations.
Reconstruction and Reporting
Standardisation of reconstruction settings is necessary in order to obtain comparable resolutions and recoveries and make SUVs interchangeable. Reconstructions must incorporate all the corrections needed for quantitative analysis, such as corrections for decay, dead time, attenuation, scatter and normalisation (correction for detector efficiencies). The following indicative settings also apply to different scanner types.. These indicative settings usually provide results that meet those QC specifications. In case these reconstructions cannot be set exactly as indicated below, settings may be set as close as possible to the ones listed below. However, note that reconstructions should be set such that they meet the multi-center QC specifications (see comments under pitfalls) 1. Siemens/CTI scanners: • (FORE+)2D OSEM with 4 iterations and 12 up to 16 subsets, • 5 mm FWHM Gaussian reconstruction filter in all directions • model-based scatter correction (default) • attenuation correction (CT or transmission source) • normalisation correction • matrix size of 128x128 up to 256x256 • zoom =1.0 2. GE scanners: • Default reconstruction setting (OSEM with 2 iterations and 30 subsets (2D scans) or FORE + 2D OSEM 5 iterations with 32 subsets (3D scans) • 5 mm FWHM Gaussian filter in all directions • model-based scatter correction • attenuation correction (CT or transmission source) • normalisation correction • matrix size 128x128 up to 256x256
3. Philips scanners (Gemini, Gemini ToF): • There are two options: (1) Gemini TF: LOR-TF-RAMLA (“Blob-OS-TF”) with ‘normal’ smoothing setting. (2) Gemini (non ToF): LOR-RAMLA using default settings (as of mid-2006). • Further default settings for options 1 and 2 (where these are adjustable): - attenuation correction (CT or transmission source) - model-based scatter correction - normalisation correction - matrix size of 144x144
Exceptions/special features Various new types of cameras are coming onto the market. It is not yet possible to specify rational dosage, acquisition and reconstruction specifications for them. Moreover, default reconstruction settings may change over time. Therefore, institutions may deviate from the recommended/prescribed dosage and acquisition protocol if it can be demonstrated that the alternative protocol provides equivalent data. This must be demonstrated by means of SNR at the level of the image, not in terms of NEC or NECR. The resolution must also match the study protocol specifications. Compliance with these requirements must be demonstrated by means of the tests described under Quality Control and inter-institution cross-calibration in addenda B to D. Calibration and activity recovery coefficients may not deviate from multi-centre standard specifications or the average results for the institutions taking part by more than 5%. These specifications are given in the addendum (chapter 14). In other words any (combination of) acquisition and reconstruction protocol and/or settings, which meet the multi-center QC specifications given in chapter 14, and especially those for the (absolute) recovery coefficients, are allowed
At present, some Philips Allegro scanners do not have a correct/accurate random scatter correction algorithm. Institutions are advised not to use these scanners for the time being if absolute SUV quantification is required (unless the reconstruction software of the scanner has been adjusted). The current scatter correction algorithm of the Philips Allegro has a less significant impact on the quality of visual assessment and applications in longitudinal studies examining tumor response on the basis of SUV ratios. Therefore, participation in multi-centre response monitoring studies is not excluded.
Detail: Reporting The reconstructed images are assessed from a computer screen. The presence or absence of abnormal FDG accumulations, especially focal accumulations, in combination with their size and intensity are evaluated. Absence of such accumulations is particularly significant if other tests have revealed findings such as anatomical abnormalities. Where necessary (see 11. Interpretation and pitfalls) the report correlates these findings to other diagnostic tests and interprets them in that context (in consultation with a radiologist where necessary) and considers them in relation to the clinical data.
Both uncorrected and attenuation-corrected images need to be assessed in order to identify any artefacts caused by contrast agents, metal implants and/or patient motion.
Criteria for visual analysis must be defined for each study protocol.
Standard uptake values are used in multi-centre studies in addition to visual assessments. SUV is a measurement of the uptake in a tumor normalised on the basis of a distribution volume. It is calculated as follows:
Actvoi (kBq/ml) SUV = Actadministered (MBq)/BW(kg)
The following calculation is applied in the case of plasma glucose correction Actvoi (kBq/ml) 5.0 SUV = x ----------------------
Actadministered (MBq)/BW(kg) Glucplasma
In these calculations, Actvoi is the activity measured in the volume of interest (see 14.Addendum A), Actadministered is the administered activity corrected for the physical decay of F18-FDG to the start of acquisition, and BW is body weight. Any alternative SUV normalisation processes (LBM, BSA) can easily be added afterwards if the patient's height, weight and gender are known (1). N.B.: the measured glucose content (Glucplasma) is normalised for an overall population average of 5.0 so that the SUVs with and without correction of glucose content are numerically practically identical (on average).
Detail: Interpretation and pitfallsAn FDG-PET scan of a fasting patient will show physiological FDG uptake mainly in the brain, urinary tract and to varying degrees in the myocardium and the colon. A high level of uptake in all skeletal muscles suggests that the patient was not fasting (obviously, this cannot be deduced from a normal serum glucose level). Detection limits naturally depend on the degree of contrast between the tumor and its immediate surroundings. It has been clearly shown that the sensitivity of FDG-PET is much lower in diabetic patients (5), though only for pancreatic cancer. There is no single detection limit for FDG-PET as it depends on many factors. The most significant of these are: histology (FDG avidity of the type of tumor), the volume of vital tumor cells, movement during acquisition (e.g. blurred signals in the case of pulmonary foci), and physiological uptake in the adjacent background. Though it is impossible to describe universal rules for detection limits, it has been demonstrated that even in the case of tumors that take up FDG in large amounts, such as melanoma, the sensitivity of FDG-PET declines very rapidly when the diameter of the tumor is less than 6 mm (6). Aspecific, non-physiological uptake is based on inflammatory processes or uptake in brown fat (neck, upper mediastinum, paravertebral region) (7). In patients who have undergone surgery, uptake therefore depends on the extent of surgery and how far the wound has healed: for example, there are few visible signs of a mediastinoscopy after ten days but a sternotomoy will remain visible for months. The pattern for bone fractures is more or less the same as has been established for skeletal scintigraphy.
Though there are no conclusive data on the optimum interval between chemotherapy and PET, an interval of at least two weeks is generally considered between the last treatment and PET. This is because of any possible effects on tumor metabolism (such as macrophage impairment) and systemic effects (such as bone marrow activation following bone marrow depression, which may or may not be caused by growth factors). The effects of growth factors (Gm-CSF) or FDG biodistribution (bone marrow!) do not last for more than two weeks after the final administration. It is assumed that the effects of radiotherapy are somewhat longer-lasting; investigation of cases of laryngeal carcinoma treated by radiation has shown that it is best to wait for about four months after the end of treatment before conducting FDG-PET. This timing fits well into this clinical context as these patients rarely develop clinical problems in the first four months after treatment.
FDG-PET is generally assessed using visual criteria (in the context of oncology, looking for a focally increased uptake that may be compatible with malignancy in the clinical context (8)). It is unclear how far semi-quantitative measurements such as SUV can contribute to the assessment, partly because of the considerable variability in the methodology used (1, 9).This Recommendation is an attempt to increase uniformity of FDG PET investigations in multi-centre studies. It is therefore also essential that the equipment used is comparable. This can be achieved by means of calibration and cross-calibration, as described in 14. Addenda B and C.
Archival Requirements for Primary Source Imaging Data
Detail: Data should be archived in DICOM 3.0 format or the current version of DICOM recommended by XXX WG YY of the XXX.
Detail: De-identification / Anonymization Schema(s) to Be Used
- Ideal: Imaging data for analysis at central laboratories should be de-identified prior to transfer. The de-identification software should be certified as fit-for-purpose by regulatory authorities at both the site of origin and site of receipt. All personal patient information that is not needed for achieving the specific aims of the trial should be removed. Pre-specified data, such as height, weight, and in some cases, sex, race, or age, may be retained if it has been approved for use by regulatory authorities. Quality assurance procedures must be performed by the recipient to verify that the images that will be submitted for analysis have been properly de-identified.
- Target: Same as ideal
- Acceptable: Data should be transferred to the "quarantine area" of a "safe harbor" for de-identification by professional research organizations or trained operators using procedures that have been certified by regulatory authorities at both the site of origin and the site of receipt. Quality assurance procedures performed by the recipient should verify that the images that will be submitted for analysis have been properly de-identified. Images that were not properly de-identified prior to receipt by the central archiving facility should be obliterated after assuring that copies conform to quality standards for patient privacy.
Detail: Archival and Transmission of Image Data Transmission of Imaging Data from Sites to Central Archive
- Ideal: electronic transmission of encrypted data over a secure network
- Target: electronic transmission with a secure file transfer protocol
- Acceptable: courier shipment of physical media containing electronic copies of the data
Note: The submission of films for digitization is not acceptable
Requirements for Local Retention of Imaging Data
Retention should conform to local law governing patient care or the requirements of the clinical trial, whichever is stricter in terms of retention times and retention format
Requirements for Central Management of Imaging Data and Imaging Metadata (e.g., the results of image analysis)
Get Clinical Computer Validation and Quality Assurance to propose a passage that can be vetted by other pharma compaies
Post-processing (i.e., anything not done on an acquisition platform that affects DICOM image data and/or pixel / voxel values)
None prior to importation into free standing image analysis software package
Detail: Image quality assessment to confirm correctness and completeness of image submissionImage set loaded to Reader (Physician performing Image Analysis) worklist Reader selects worklist task Digital image datasets within assigned timepoint opened for display
Detail: Volume of interest (VOI)Definition The following 3D-volumes (volumes of interest, VOI) may be determined: 1. 3D isocontour at 41% of the maximum pixel value with background correction (VOI41BG) 2. 3D isocontour at 50% of the maximum pixel value (VOI50) 3. 3D isocontour at 50% with background correction (VOI50BG ) 4. 3D isocontour at 70% of the maximum pixel value (VOI70) 5. 3D isocontour at 70% with background correction (VOI70BG ) and 6. Maximum pixel value (Max)
• The isocontour described in point 1 (VOI41BG) generally corresponds best with the actual dimensions of the tumor (1), but only for higher tumor-to-background values and homogenous backgrounds. In practice, however, this VOI seldom results in useful tumor definition because of noise, inhomogenities in tumor and background, and sometimes low tumor-to-background ratios (low contrast between tumor and background). In this case, the VOI based on a higher isocontour value should be chosen. N.B. Other tumor segmentation methods have been described for tumor volumetry in literature. These, however, are not (yet) routinely used for determining SUVs and have to be further evaluated. • The VOIs described in (1) to (6) form a series of increasing isocontour values, which result in smaller VOIs. In longitudinal studies (response monitoring), the VOI should be chosen that results in useful tumor definitions in all PET scans. This is based on the fact that largest VOI are most precise and thus most suitable (most reproducible) for measuring response [2-4, 6]. A reasonable alternative in multi-centre trials is therefore to always opt for a 70% isocontour VOI and/or the maximum pixel value.
VOIs can be created semi-automatically using the “NEDPAS” software (VUmc, available on request, firstname.lastname@example.org). The program has I/O for ECAT7 format files and DICOM provided this complies with Siemens/GE/Philips conformance statements. Other formats and/or DICOM dialects are not accepted and/or cannot be read. The program generates a report which gives the SUV for all VOIs. If it is not possible to create a useful/realistic VOI using the automatic method due to the presence of high background near to the lesion, the region of interest should be drawn manually (3D, using standard scanner software) with reporting of the maximum SUV value only. Local (institution’s own) software may be used of course, provided it is compatible with the VOI method required for multi-centre study. Always check automatically generated VOIs visually. SUVs are calculated with and without correction for plasma glucose. Pitfalls: VOIs are generated semi-automatically, but it is often not possible to generate a reliable VOI if there is a high background or an area of high uptake (bladder, heart) close to/adjacent to the lesion, or if there is low uptake in the lesion. Semi-automatically generated VOIs must therefore be checked visually. If the VOIs are not reliable and/or do not correspond visually with the lesion, only the maximum SUV based on a manually generated VOI should be used for reporting.
Detail: Timepoint exams defined by charter re: exam types and dates of inclusion Reader performs review of imaging exams within defined timepoint For standard oncology trial, Baseline only timepoint is reviewed first Reader iterates with software application to identify lesions and perform lesion analysis Target lesion identification and analysis Non-target lesion identification +/- analysis (or comments) Lesion analysis performed by providing measurement object as spearate file Label lesion Capture metadata generated from measurement object Repeat Target +/- Non-target lesion identification/analysis as defined by imaging charter Close subtask
Detail: @Open next timepoint for given subject Review next ‘on-study’ timepoint with prior timepoints Perform lesion analysis on current timepoint to correspond to lesions identified before Identify presence/absence of New lesions Derive change metrics for lesions +/- lesion group (e.g. target lesions) Determine response assessment as defined by imaging charter (Evaluate covariates and other possible sources of technical variation for assessment of affect likelihood upon response assessment)* Close subtask Continue same workflow (from @) until all timepoints reviewed for given subject
Radiation Dose and Safety Considerations
It is recognized that X-ray CT uses ionizing radiation and this poses some small, but non-zero risk to the patients in any clinical trial. The radiation dose to the subjects in any trial should consider the age and disease status (e.g. known disease or screening populations) of these subjects as well as the goals of the clinical trial. These should inform the tradeoffs between desired image quality and radiation dose necessary to achieve the goals of the clinical trial.
Imaging Agent Dose and Safety Considerations
Imaging Hardware-specific Safety Considerations
• Propose that vendors make QC results (only pass/fail/warning) and calibration results (mean, variance) readily accessible as external files or ideally incorporated in the public DICOM tags of each patient study • PET/CT Quality Control requirements for accurate and precise SUV quantification • CT scanner calibration and QC: • Rationale: Accurate CT Hounsfield units are required for proper display of anatomic images, correction for PET attenuation, and response assessment based on tissue density changes tumor size measurements (e.g. RECIST Chesson, and Choi criteria). • Limitations: All vendors provide acceptable daily calibration and quality control checks but the results are not documented and readily accessible. • Proposed Solution: Daily QC results of the mean and variance of a water-equivalent phantom should be easily exported as an external file and/or incorporated into public DICOM tags associated with each patient PET/CT for that day. • PET Scanner calibration and QC: • Rationale: Accurate radioactivity measurements are required for quantitative PET imaging and SUV analysis. The results of quality control tests need to be documented to demonstrate acceptable scanner performance in clinical trials and clinical practice. • Limitations: PET scanner quality control procedures are generally acceptable but there is insufficient documentation. PET scanner qualification is typically performed at the initiation of clinical trials but there is not a readily available means to verify and document acceptable performance throughout the trial. • Proposed Solution: The results of each vendor’s scanner QC should be exportable as an external file and ideally incorporated in public DICOM tags with a status of pass/fail/warning. A uniform cylinder calibration test should be performed as part of the daily QC (analogous to the CT HU calibration) and documented in an external file and/or the public DICOM tags of each patient study.
Quality Control and Inter-institution Cross-Calibration Both physiological and physical factors influence the accuracy and reproducibility of ‘standard uptake values’ (SUV) in oncological FDG-PET studies. . Variations in PET camera calibration, image reconstruction and data analysis and/or settings can have more than a 50% effect on the measured SUV [2-4]. The use of SUV in multi-centre oncological PET studies therefore requires an inter-institution calibration procedure in order to facilitate the interchangeability of SUVs between institutions. It is also important that all participating institutions use methodology which is as similar as possible. In order to ensure the interchangeability of SUVs, a minimum set of quality control procedures must be carried out, such as: 1. daily quality control 2. calibration/cross-calibration of PET or PET-CT camera with the institution’s own or against another dose calibrator which is generally used to determine patient specific FDG doses. 3. inter-institution cross-calibration and determining ‘activity recovery coefficients’. Brief summary of the quality control procedures:
1 Daily quality control (Daily QC) The aim of daily quality control is to determine whether the PET or PET-CT camera is functioning well; in other words to establish detector failure and/or electronic drift. Most commercial systems are equipped with an automatic or semi-automatic procedure for performing daily quality controls. For some scanners, the daily quality control includes tuning of hardware and/or settings. Thus both the procedure and its name may be different between various scanners. In all cases all daily quality control measures and/or daily setup/tuning measurements should be performed according the manufacturer’s specifications. Users should check whether the daily quality control meets the specifications or passed the test correctly.
2 Calibration QC and cross-calibration of PET and/or PET-CT cameras The aim of calibration and cross-calibration is to determine the correct and direct calibration of a PET or PET-CT camera with the institution’s own or against another dose calibrator which is used to determine patient-specific FDG doses (5). If these doses are ordered directly from and supplied by a pharmaceutical company, cross-calibration of the PET camera should be carried out using a calibration sample supplied by that company (i.e. the customer should order an FDG dose of about 70 MBq, see below, as if it concerns a patient dose). Remember that cross-calibration must not be confused with normal calibration. Cross-calibration is a direct, relative calibration between the used (or institution’s own) calibrator and the PET camera, and therefore provides information about possible calibration discrepancies between the PET camera and the dose calibrator, which is more essential for correct SUV quantification than the individual calibrations themselves. There may still be differences (of up to 15%) in the cross-calibration between PET camera and dose calibrator due to the fact that individual calibrations of the dose calibrator and the PET camera (usually carried out by the manufacturer) are performed using different calibration sources and procedures, and by different companies and/or persons. This explains the importance of a direct cross-calibration between the dose calibrator used and the PET camera used.
In short, the procedure is as follows: A syringe is filled with approximately 70 ± 10 MBq of F18-FDG solution and is re-measured in a calibrated dose calibrator (or the syringe is ordered from the pharmaceutical company). The F18-FDG is then introduced into a calibration phantom filled with an exact volume (< 1 %) of water, which results in a solution containing an exactly known activity concentration (Bq/cc). Homogenisation of the F18-FDG in the phantom should be achieved by leaving an air bubble of approximately 10-20 ml within the phantom and subsequently shaking/mixing the phantom for a short period of time (10 minutes or more). If the institution has a calibrated well counter, three samples of approximately 0.5 ml should be taken from the calibration phantom solution using a pipette. The exact weight/volume of the samples should be determined before placing the samples in the well counter. Emission scans of the calibration phantom are performed with the PET or PET-CT camera using the recommended whole body acquisition protocol/procedure (including multibed acquisitions, see appendix). Once the activity has decayed (after an interval of 10 hours or more), a transmission scan is performed without moving the phantom from its position in the scanner. For PET-CT cameras on which attenuation correction is performed using a low dose CT-scan (CT-AC), the CT-AC scan can be carried out either directly before or after the emission scan.
Emission scans are reconstructed in accordance with the recommended reconstruction parameters as described in paragraph 9, Camera and computer. VOI analysis is performed in order to determine the average volumetric concentration of activity within the phantom as measured by the PET camera. Cross-calibration factors between the PET or PET-CT camera and dose calibrator and well counters can then be derived directly. Once the cross-calibration procedure has been completed, conversion factors will be known with which the counts/measurements for different equipment can be synchronised. N.B.: The cross-calibration factor between the PET camera and dose calibrator should be equal to 1.0 (< 5%). A ‘standard operating procedure’ (SOP) is described in 14. Addendum C. Software and/or processing programs, provided by the Vumc, may be used (on request available, email@example.com). 3 Image quality and recovery coefficients (IQRC) Although a correct cross-calibration is guaranteed using the quality control procedure described above, differences in SUV quantification may still occur in multi-centre trials as a result of differences in the reconstruction and data analysis methodology used by participating institutions. In particular, differences in the final image reconstruction (i.e. following reconstruction, including all effects due to filters and pixel size settings etc.) have, depending on the shape of the tumor, a significant effect on the SUV result for smaller (< 5 cm diameter) tumors. It is therefore important to determine the accuracy of the SUV using a standardised ‘anthropomorphic’ phantom containing spheres (tumors) of varying sizes. Phantoms such as these enable us to verify SUV quantification under clinically relevant conditions. The aim of the IQRC quality control procedure is:
1. to determine/check the correctness of a calibration and quantification using a non-standard (calibration) phantom 2. to measure ‘activity concentration recovery coefficients’ as a function of sphere (tumor) size.
The IQRC quality control procedure is carried out in accordance with the ‘image quality, accuracy of attenuation and scatter corrections’ procedure described in the NEMA Standards Publication NU 2-2001, “Performance measurements of positron emission tomographs”. VOIs are defined manually according to this procedure. However, it is known that automatic definition of 3D volumes of interest (VOI) based on isocontours using fixed percentages results in a higher SUV accuracy and precision than those determined using manually defined ROIs or VOIs (2,3,6). Therefore, 3D-VOIs are also determined using an automatic VOI method such as described in paragraph 14, Addendum A:
(1) 3D isocontour at 41% with background correction (VOI41BG) (2) 3D isocontour at 50% of the maximum pixel value (VOI50) (3) 3D isocontour at 50% with background correction (VOI50BG) (4) 3D isocontour at 70% of the maximum pixel value (VOI70) (5) D isocontour at 70% with background correction (VOI70BG ), and (6) Maximum pixel value (Max) The procedure for making this VOI is as follows: Firstly, the location of the pixel with the maximum SUV in the tumor must be determined (manually or semi-automatically). Secondly, a 3D-VOI is generated automatically based on the maximum SUV/pixel value and its location with a 3D ‘region growing’ algorithm in which all pixels/voxels above the defined threshold limit are included (mentioned in points 1-6). In order to implement/apply this procedure, the VUmc will provide software with which this VOI can be automatically generated, and this phantom test can be analysed. (Software is freely available on request through firstname.lastname@example.org). Once a VOI has been generated for each sphere, the average concentration of activity (or SUV) for the sphere can also be determined. The average VOI activity concentration value measured is then normalised with the actual concentration of activity in the spheres, which indicates the ‘activity concentration recovery coefficient’ per sphere (i.e., the ratio of the measured and actual concentration of activity as a function of sphere size). The ‘recovery coefficient’ is finally defined as a function of sphere size and VOI definition. A standard operating procedure is presented in 14. Addendum D. The measured activity concentration recovery coefficients must meet certain specifications which are given below. These specifications are based on recovery coefficients measured according this procotol on various PET and PET/CT scanners (Siemens, Philips, GE). NEDPAS specifications for activity concentration recovery coefficients (RC) measured according the Image Quality QC procedure (addendum 14A,D). Specifications are given for recovery coefficients obtained using VOI50BG and the maximum pixel value only.
RC specification for VOI50BG Sphere volume (ml) Expected RC Maximal RC Minimal RC 26.52 0.77 0.83 0.71 11.49 0.73 0.79 0.67 5.57 0.66 0.73 0.59 2.57 0.60 0.68 0.53 1.15 0.45 0.52 0.38 0.52 0.30 0.35 0.25
RC specifications for maximum pixel value Sphere volume (ml) Expected RC Maximal RC Minimal RC 26.52 0.98 1.08 0.88 11.49 0.95 1.05 0.85 5.57 0.89 1.01 0.77 2.57 0.84 0.94 0.75 1.15 0.63 0.74 0.51 0.52 0.38 0.46 0.29
Minimum frequency of quality control procedures: Procedure: Frequency: Daily QC Daily Cross-calibration At least 1x per 3 months and always immediately following software and hardware revisions/upgrades and immediately following new setups/normalisations. IQRC Once per institution participating in a multi-centre trial and always following reconstruction/scanner software adjustments (especially adjustments to the reconstruction and/or data analysis (region of interest) software/hardware). All relevant data, including reconstructed images and fully completed SOPs, should be made available to the central data reviewing centre. Data must be delivered in ECAT7 or DICOM format. DICOM files must comply with the scanner manufacturer’s ‘conformance statement’ on DICOM specifications. Each institution is responsible for ensuring these specifications are met.
C. Calibration QC of PET and information required for CRF and Standard Operating Procedure for multi-centre studies 1. Requirements • Approximately 70 MBq F18-FDG • Calibration phantom (cylindrical) with exact known volume • Date • Location/hospital • Performed by • Scanner (manufacturer/type) • Volume of calibration phantom (ml)
2. Preparation • Draw up approximately 70 MBq of 18F and fill up to 4.5-5.5 cc with water, then measure the activity using the dose calibrator. N.B.: The activity required depends on the time that PET scanner measurements are to start. The dose must therefore be adjusted if necessary so that the phantom contains approximately 50-70 MBq at the time of the calibration measurement. Alternatively, this syringe can be ordered according specifications of this SOP from a pharmaceutical company. • Activity in syringe = MBq at (time) hrs. • A dose calibration measurement label may be attached below (if available).
• Remove approximately 10 ml water from the calibration phantom (which has been completely filled with water) and then introduce the F18-FDG from the syringe. Flush the syringe thoroughly. An air bubble of approximately 5 cc will therefore be left in the phantom for homogenisation purposes! • Shake the calibration phantom well.
3. Well counter (optional; only if specified in study) If the department has a well counter that is calibrated for F18-FDG: • Take 3x 0.5 ml samples from the phantom using a pipette. Before taking the sample, place the counting tube on the scales and set these to 0. Then take 0.5 ml of the phantom solution using a pipette and weigh the counting tube containing the 0.5 ml sample.
• Record the net weight of the samples in the space below: Sample 1: gr Sample 2: gr Sample 3: gr • Count the activity in the samples using the well counter(s) (30 or 60 seconds counting time is sufficient). Record the number of CPM (corrected for dead time) measured by the channel calibrated for F18. Make a note of the start time for counting. CPM sample 1: cpm CPM sample 2: cpm CPM sample 3: cpm Start time for counting (hh:mm:ss)=
4. PET scans • Place the phantom in the scanner. Perform an emission scan using the acquisition parameters described in paragraphs 7, 8 and 9 (note that the emission scan will now take a minimum of 10 min for PET/CT scanners and 30 min for PET only scanners) A transmission scan of at least 10 minutes will be made after a 10-hour interval, or a CT-AC using standard settings can be performed immediately before or after emission scanning. Emission scans should be performed in both 2D and 3D (if this is possible). • Record start times for scans, please read from console and later check in file header (if possible): Acquisition time of 2D scan = Acquisition time of 3D scan =
5. Reconstructions ‘Quantitative’ reconstruction parameters must be used as specified in paragraph 9; this applies for all scanners.
6. Archiving for multi-centre studies and analysis According to the relevant study protocol. Please note: The reconstructed images must be supplied in DICOM format. DICOM files must comply with the relevant scanner manufacturer’s ‘DICOM conformance statement’.
7. Pitfalls • Ensure that clocks are synchronised in all working areas (in other words, scanner area, on the PET or PET/CT camera itself, on the PET or PET/CT camera computers, in the hot lab or analysis lab and on/nearby/around the dose calibrator being used and/or on the dose calibrator’s computer). Non-synchronised clocks and incorrectly reported times will result directly in calibration errors! A 15-minute mismatch, for example, results in an avoidable calibration error of 10%. • Not flushing the syringe adequately when introducing the activity into the phantom means that not all the activity in the syringe is transferred to the phantom and will result in incorrect interpretation of the results. • Check whether the activity in the syringe has been measured with or without a clean needle. Particularly where small volumes are concerned, a large proportion of the activity can be present in the needle during the dose measurement. If the needle is subsequently exchanged for a ‘clean’ needle, the dose measurement will no longer correspond with the actual net activity in the syringe. Some types of scanners require the correct phantom weight (= net volume), the dose and the time to be filled in when the scan protocols are applied. If this information is omitted, the reconstruction software will not implement correction factors for quantification. Unfortunately, this may also apply to human studies!. Be aware that dose is specified at dose calibration time, which is in practice not necessarily equal to injection time or start of the PET/CT study. Apply necessary corrections for decay, if needed or use dose calibration time.
D CRF/SOP Image Quality and Activity Concentration Recovery Coefficient PET (for multi-center studies)
1. Requirements • F18 activity (volume not important) in 2 ml syringes • 1x 10 MBq, 1x 20 MBq (record the volume and clearly record the time of calibration) • 1 measuring cup of exactly 500 ml • Felt-tip pen • NEMA NU2-2001 (section 7) Image Quality phantom with exact dimensions/volumes.
2. Preparation Stock solution for the spheres: • Fill the measuring cup with exactly 500 ml • Add the 10 MBq F18-FDG, record the exact dose below, plus the time of calibration and volume in the syringe: Activity = MBq at (hh:mm:ss) Volume (of solution of activity in syringe) = (ml) • Stick the dose calibration measurement label (if available) in the space below • Homogenise the solution thus made • Fill all the spheres from the NEMA NU2-2002 image quality phantom with stock solution. Note that this deviates from NEMA NU 2-2001 (section 7) procedures!
Filling the Image Quality Phantom • Introduce the spheres into the NEMA NU 2-2001 phantom (if not fixed within the phantom) • Fill the background compartment of the phantom completely with water • Remove 20-30 cc water from the phantom • Empty the syringe containing the 20 MBq into the background (flush well so that all the activity is introduced into the phantom) • Homogenise the phantom
3. Scans • Make standard quantitative whole body scans (2 bed positions) of the phantom according to the specifications in paragraphs 7 to 9; emission scans of at least 10 mins/bed position, however, are required for this test. • Record the scan times: mins emission (3 mins transmission, if no CT-AC). • Record T = 0 (start time of the scans)
4. Reconstructions In accordance with specifications given in paragraph 9.
5. Analysis The following data can be determined using the “Image Quality QA” program: • The average concentration of background activity in the reconstructed PET images using several VOIs. The accuracy of the PET scanner cross-calibration for the image quality phantom can be derived from this. • The average pixel size in the ‘scatter’ insert, to check the scatter correction. • The average concentration of activity in the spheres based on isocontours (described in the protocol) and the resulting activity recovery coefficients as a function of sphere size. • The accuracy of the calibration and activity recovery coefficients as a function of sphere size are used for reporting and should meet the specification given previously. • Explanatory note: Dose calibrations are performed at the study centre itself or at the FDG supplier’s site. N.B.: Supplies of FDG may have smaller volumes.
6. Archiving/multi-centre analysis In accordance with study protocol. Please note: • Relevant data (completed CRF and reconstructed images) must be made available to the central data analysis centre for multi-centre quantification monitoring. • The reconstructed images must be supplied in ECAT7 or DICOM format. • DICOM files must comply with the relevant scanner manufacturer’s ‘DICOM conformance statement’. Each institution is responsible for ensuring this requirement is met.
Non-conformance to the specified data format makes quantification of the study impossible, resulting in the institution being excluded from any further participation in the study. In mutual consultation, it may be possible to find an alternative solution.
7. Pitfalls • Ensure that clocks are synchronised in all working areas (in other words, scanner area, on the PET camera itself, on the PET camera computers, in the hot lab or analysis lab and on/nearby/around the dose calibrator being used and/or on the dose calibrator’s computer). Non-synchronised clocks and incorrectly reported times result directly in calibration errors! • Not flushing the syringe adequately when introducing the activity into the phantom means that not all the activity in the syringe is transferred to the phantom and will result in incorrect interpretation of the results.
• Check whether the activity in the syringe has been measured with or without a clean needle. Particularly where small volumes are concerned, a large proportion of the activity can be left behind in the needle during the dose measurement. If the needle is subsequently exchanged for a ‘clean’ needle, the dose measurement will no longer correspond with the actual net activity in the syringe. • Some types of scanners require the correct phantom weight (= volume), the dose and the time to be filled in while the scan protocols are being applied. If this information is omitted, the reconstruction software will not implement correction factors for quantification. Unfortunately, this also applies to human studies!
8. Other recommendations If the department has a calibrated well counter available (see calibration procedure), this is the tool of preference with which to determine/verify the exact concentration of activity in the spheres and in the background of the phantom.
Detail: Subject preparation
Detail: Imaging agent dose calculation
Detail: Imaging agent-related
Detail: Image data acquisition-related
Detail: Inherent image data reconstruction / processing
Detail: Image analysis and interpretation
Detail: Site selection and Quality Control
The FDG-PET Committee is deeply grateful for the remarkable support and technical assistance provided by the staff of the Radiological Society of North America, including Susan Anderson, Linda Bresolin, Joseph Koudelik, and Fiona Miller.
Suggestion for attribution: Everybody who ever made a call and/or wants to go on record as endorsing the work should be included. A list of 1,000 names doesn't seem as though it would be too long. This is going to be published on the web. If preferred, then we can list the "regulars" as "core members" or some other euphemism.