| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Full paper |
1 Northern Ireland Regional Medical Physics Agency, Royal Victoria Hospital, Belfast, UK, 2 U650 INSERM, LaTIM, Brest, France
 |
Abstract |
|---|
 Top Abstract IntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
Beyond this, gating systems are being introduced which permit the radiation to be delivered in synchrony with physiological motion such as respiration [2–5]. This is then coupled with the development of treatment plans and target volumes in equivalent physiological states requiring 4D plans including space and time. The raison d'etre for these technological advances is that a more accurate delivery of radiation to appropriate target volumes will lead to less damage to "normal" tissue whilst permitting higher, more targeted and non-uniform doses to be applied to the "diseased" tissue volume. In order to fully exploit these delivery system developments, the treatment volume planning process needs accurate information about the spatial location and extent of the target volume. In addition, information about intrafractional motion of the target and the impact of this on the delivery and treatment plan is required.
It is necessary to define what is meant by radiotherapy treatment planning within the context of this review. Treatment planning may be regarded as encompassing the process that the oncologist undertakes when a patient is referred into their care. This includes gaining accurate diagnostic and staging information from a range of tests, planning the overall treatment as well as defining the target volume and specifying the radiation treatment portals. A more specific definition of treatment planning is to consider only the process of target delineation and the subsequent determination of the treatment beam shapes and orientation. In this context, it is assumed that the oncologist is fully aware of all the staging issues such that the role of the PET image is now to support the delineation of the tumour volume. For the purposes of this review, treatment planning shall be considered to be the process of target volume delineation.
Radiotherapy treatment plans are developed using X-ray CT images. These provide data about the X-ray attenuation characteristics of the patient as well as identifying structures and organs for the specification and delineation of treatment volumes. It is clearly understood that this property may not reflect the underlying cellular and biochemical processes and that for the treatment to be effective it must also target those tissues where function rather than anatomy is abnormal. This has therefore led to the investigations into and the introduction of imaging methodologies which can supplement the purely anatomical information of X-ray imaging by providing data regarding molecular function. MRI can provide functional information, although limited in extent (see Khoo et al, and Payne and Leach in this issue). However, techniques based on the radioactive tracer method offer sensitivities to biochemical processes which are 103–106 higher than MRI, depending on the radioactive tracer and the detection technology. Positron emission tomography (PET) is one such technology. Current systems combine a PET detector with a CT scanner and can provide geometrically aligned anatomical data (the basis of the current treatment planning process) and functional information at the molecular level [6]. These functional data can be readily introduced into the planning process using current generation radiotherapy treatment planning systems. Oncologists have used PET images in the diagnostic and staging process, especially to make decisions about the inclusion of suspected nodal disease. These decisions can result in significant changes in treatment volumes when compared with a process where PET data has not been used [7]. The accurate staging of the disease using PET is known to be essential for the appropriate management of the patient [8].
To date, PET images have been applied in radiotherapy planning primarily for non-small-cell lung cancer (NSCLC) [2, 9–11]. However, studies into the use of PET images for other sites, such as oesophagus [12] and head and neck cancer [13] have also been published. All these studies have shown that the inclusion of PET data in the treatment planning process modified the target volumes in a significant number of patients. What is less clear is whether these changes are primarily due to the inclusion of previously unsuspected nodes in the target volume, as the patients had generally not had a staging PET scan in addition to the scan used for treatment planning purposes. It is well known that there may be high interobserver variability between gross tumour volumes (GTVs) defined using planning CT images. Several studies have shown that the inclusion of PET data reduces this interobserver variability [9, 11, 14]. Although this is a welcomed outcome, it does not necessarily mean that the volumes are being defined any more accurately [15].
From the above published studies, it can be seen that volumes can be delineated and treatment options changed based on the additional information. However, such an approach can only be justified if it is possible to adequately answer a number of key questions: (i) What does the molecular image represent, how specific is the functional signal and how sensitive to the disease process? (ii) What does the uptake of the radiotracer mean? (iii) How are the PET data best acquired? Do the limitations imposed on the acquisition of PET/CT data impact upon clinical utility and outcomes? (iv) What criteria should be used to delineate target volumes on PET images? (v) Can the functional data acquired be synchronised to a physiological signal such as respiration? What will be the impact on volume delineation and treatment delivery? (vi) Does the incorporation of functional data change clinical outcomes by improving quality of life or by improving survival?
The following sections review the current status of the answers to these questions and outline some of the work which is being undertaken to evaluate the impact of the unknowns on the introduction of PET into the radiotherapy planning process.
|
The PET image: its acquisition and interpretation |
|---|
|
TopAbstractIntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
Although FDG is currently the predominant tracer used in oncology studies and its use will be the focus of this review, tracers mapping different metabolic processes may also have a role. Amino acid metabolism may prove to offer a useful non-specific index of tumour viability, especially in the brain where there are high rates of glucose metabolism in normal nervous tissue. This process can be probed with 11C-Methionine or 18F labelled tyrosine. Thymidine has also been proposed as a marker of cell proliferation based upon its incorporation into DNA [17]. A thymidine analogue, 3'-deoxy-3'-[18F]fluorothymidine (FLT) can now be routinely manufactured and is entering clinical trials. Of growing interest is the possibility of differentiating tissue or cell types with images which provide data related to clonogen densities. Currently, tracers to image hypoxia, angiogenesis and apoptosis are under development. For hypoxia imaging, the most widely studied tracer is a misonidazole derivative (18F-MISO) together with a copper labelled compound, diacctyl-bis(N(4)-methylthiosemicarbazone (ATSM). The use of this latter compound to guide IMRT treatment planning has already been reported [18]. These and many other compounds will continue to be developed and each will need validation as an imaging and treatment guidance tool.
PET image interpretation
How should the PET data be used in the radiotherapy treatment process and, in particular, in the delineation of the target volume? A recent editorial by Gregoire [19] highlights the issues in relation to false positive and false negative rates for detection of disease. It must be recognized that no single modality has 100% sensitivity and specificity for disease. A gold standard does thus not presently exist. The definition of treatment volumes, and hence treatments based upon a single diagnostic modality, may fail through the exclusion of diseased tissue not detected. Historical clinical trial data [20] shows that in the majority of lung cancer patients disease progression is due to recurrence local to the treatment volumes. This may be due to insufficient radiation dose being delivered to the tumour, with the prescribed dose generally being limited by the tolerance of normal tissue within the treatment field. Alternatively, it may be due to a geographical miss of the real tumour. Improved knowledge of the actual tumour size, shape and location may permit smaller treatment fields allowing increased radiation doses to be prescribed. Increasing volumes will minimize the risk of geographical misses. The delineation of volumes must be based on all the diagnostic information and knowledge available of the anatomy and physiology of disease. A new modality such as PET must be integrated into this knowledge base with decisions on treatment modification based upon probabilities of false positive and negative data within particular structures and locations [19]. Simple but logical approaches are currently being applied in clinical situations [21]. The following argument might be applied to a protocol for target volume modification in the treatment of lung cancer based upon CT data only; (a) if the PET scan is positive then volumes will be added into the treatment plan. This will result in non-neoplastic tissue being incorporated into the treatment volume on a number of occasions, but will limit the non-irradiation of neoplastic cells. The effectiveness of this strategy will be dependent on the increased sensitivity of the new modality (PET imaging) compared with CT. (b) If the PET is negative then volumes should be removed from the planned treatment volume. This approach requires a more specific PET image than can be obtained with CT. It will potentially result in neoplastic tissue being excluded from the treatment volume, but will minimize the volume of normal tissue irradiated with a therapeutic dose. The probability of a correct interpretation of a diagnostic image is dependent on a number of factors including: the spatial resolution of the modality; the signal to noise ratio of the image formation process; the affinity of the tracer or contrast material for neoplastic tissues; confounding processes such as inflammation, as well as the criteria used for image interpretation; and last, but not least, the interpretive skills of the reporting clinician and radiation oncologist.
PET and PET/CT images for treatment planning
Images for radiotherapy treatment planning (RTP) purposes must be acquired differently from those for use in diagnosis and staging of patients. The standard imaging technique used in radiotherapy planning is X-ray CT as it provides both good anatomical detail for defining target volumes and the electron density data required for dose calculations. Any other images for use in RTP normally have to be registered to a planning CT scan. Images for use in RTP must be acquired with the patient in the treatment position, on a flat couch top and with the use of appropriate immobilization devices. Patient set-up must be carried out by therapy radiographers. The initial set-up of the patient in the treatment position may take some time. The internal target volume defined from the images must be related to the surface of the patient for subsequent set-up of the patient for treatment simulation and delivery. This is achieved using markers which are visible in the image to be located at skin positions which are semi-permanently marked. There are a number of different ways in which PET images can be used for treatment planning purposes. PET images could be available purely as a diagnostic aid; PET and CT images from separate scanners can be registered in software; PET-CT images from a combined scanner could be registered to a planning CT scan; or a planning PET-CT scan could be carried out on a combined scanner.
Initial studies [22] of the use of PET images in treatment planning used PET and CT images which had been acquired on different scanners and registered afterwards. In general, these studies had been acquired with the patient in the treatment position. Software fusion of the PET and CT images has the advantage that the acquisition of the PET data does not necessarily need to be conducted in the treatment position and might be obtained from the diagnostic scan. The disadvantage is that the simulation scans and the PET scans will be acquired at different times, possibly on different couch tops (flat top compared with a concave top) and potentially with the patient in different positions. In a review of "organ motion and its management" Langen and Jones [23] refer to such position-related organ motion and the potential problems associated with imaging and treating with the patient in different positions. These can be caused in part by differences in deflection of the imaging and treatment couches as well as setup errors. If images are acquired in this way, for sites other than rigid organs such as the skull and brain, there will be a requirement for non-rigid body data transforms to register the two datasets [24]. Such image registration techniques are known to result in residual misalignment errors, the clinical significance of which remain unknown. Another option is to use the CT scan obtained for the attenuation correction of the PET images as the basis for registration with a separate radiotherapy planning CT scan. This has the advantage that the image registration is CT to CT and hence potentially has smaller errors associated with it.
More recently, a number of studies have been published which have used combined PET-CT scanners [14]. Although the use of PET-CT scanners to provide a hardware solution to image registration has the advantage of reducing uncertainties in patient positioning, there is still an on-going debate about the efficacy of such integrated scanners [25] for radiation oncology.
If a PET or PET-CT scanner is to be used directly for acquiring radiotherapy planning scans then a number of issues need to be considered. The introduction of any new imaging modality for RTP requires the adaptation of scanning procedures for use with that modality. Close liaison between Oncology and Nuclear Medicine staff will be required to develop these protocols. Acquiring the PET/CT dataset in the treatment position requires the direct involvement of radiotherapy treatment radiographers to ensure that protocols are consistent across the planning and treatment processes. Set-up of the patient for treatment is a time consuming process and, if undertaken for PET scanning, it will be necessary to develop procedures that minimize contact between the therapy radiographers and the patient, after the patient has been injected with the radioactive tracer, in order to minimize radiation dose to staff [26, 27]. The patient will have to be positioned so that they fit through the bore of the PET-CT scanner. External wall or floor mounted positioning lasers may need to be fitted in the PET-CT scanning room. PET scans take considerably longer than CT scans, so the patient will need to maintain the treatment position for longer than for CT planning scans, simulation or treatment.
Decisions will have to be made about whether dedicated PET-CT planning scans are going to be acquired, or whether whole body diagnostic scans are to be acquired in the treatment position and used for both staging and RTP. A potentially significant number of these patients will, however, not progress to radiotherapy treatment due to the outcome of the staging scan. PET scanning times are relatively long, depending on the acquisition protocol, and patient compliance may be an issue in terms of maintaining the position throughout the scan. The time taken to acquire the PET component of a PET-CT dataset depends on the number of fields of view required. A planning PET scan might require, for example in the case of the lung, two fields of view resulting in a minimum scan time of 8–10 min on current systems. Alternatively, the process could be integrated into the diagnostic work-up to preclude the need for a second radiotherapy planning study. In this scenario, the patient would need to maintain the treatment position for 30–35 min.
|
Target volume delineation for radiotherapy planning |
|---|
|
TopAbstractIntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
Thresholding is the most widely used method to determine volumes automatically from PET images [30–32]. It is well known in nuclear medicine that the selection of the threshold depends on the lesion size, shape and contrast. Figure 1
illustrates the effect of different thresholding and analysis techniques on segmented volumes. It has been shown that when a priori knowledge of the size, shape and contrast of the lesion exist, threshold levels can be specified for accurate evaluation of the tumour volume. Most algorithms have been derived from the study of spheres of various sizes with differing levels of background signal. All groups have recognized the need to take account of the lesion to background ratio in defining volumes. All have recognized that the volume calculation is in error in small volumes (typically <4–8 ml depending on the resolution of the system). However, little work has been published on the effect of noise levels within the images or on the validation of volume calculations from non-spherical objects. There has also been little validation of the use of PET imaging to delineate the GTV through the use of surgical samples [33, 34]. A number of groups have proceeded to implement volume definitions into the radiotherapy planning process based on this phantom validation work and used contours based upon the maximum in the lesion of interest with values ranging from 30% to 50%. Paulino et al [35] used a threshold of 50% in head and neck tumours and reported that 75% of GTV definitions were smaller using the PET image when compared with the CT derived volumes. Such a consistent finding would suggest an inappropriate choice of contour value. Other factors such as lesion volumes, reconstruction software, tracer sensitivity and specificity could all contribute to these findings. Other groups have chosen to use a threshold based upon the standardized uptake values (SUV) rather than upon a threshold based on maximum intensity [36]. What is not in doubt from these publications is that different GTVs have been obtained from the PET and the CT data and that treatment volumes have been modified. Bradley et al [7] reported a change in volume in 50% of patients based upon GTVs. The group continued to add margins to these data to provide the final PTV. In the chest for lymphoma, Lee et al [37] based volumes upon manual contours derived from PET images which had been thresholded to provide a PET defined lung volume which "matched" the CT lung volume.
|
In defining target volumes in the thorax and upper abdomen, respiratory motion must be taken into account. As described above, margins are added to the GTV to take this into account. There is also considerable interest in applying respiratory gating to both the planning PET-CT scans and the radiotherapy treatment.
|
Respiratory gating for PET/CT data in radiotherapy planning |
|---|
|
TopAbstractIntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
The definition of a PTV clearly remains a complex task and techniques for the definition of a GTV may have a limited impact on the final definition of a PTV, especially where significant physiological motion is known to occur. These uncertainties have led to the investigation of diagnostic and treatment methodologies which measure physiological motion and incorporate the data into the treatment plan and the delivery system.
The solutions that have been proposed to date for taking into account the effects of respiratory motion concentrate on the acquisition of respiration synchronised PET and CT datasets. The use of breath hold protocols has been used as a means of improving registration between the PET and CT [43]. However, these will not aid in the delineation of target volumes as the radiotherapy treatment will be delivered over a few minutes. There is a lot of interest in the use of respiratory gating for both the PET-CT image acquisition and the treatment. Several studies have been carried out to investigate the feasibility of respiratory gating of PET of the upper chest and abdomen [44–46] and also to quantify the impact of respiratory motion on the underestimation of lesion activity [47]. Different detector systems have been proposed, including a transducer or an impedance electrocardiograph (ECG) monitor measuring changes in abdominal or thoracic circumference [48], a thermistor measuring the temperature of circulating air during patient respiration [45, 46], a spirometer measuring respiratory flow [49], the Varian Real Time position management (RPM, Varian Medical Systems, Palo Alto, CA) [47, 48], or the Polaris system tracking the displacement of infrared reflective markers in the patient chest [47]. An alternative approach to gating is to use an image derived respiratory signal through the acquisition of dynamic datasets [51, 52] or list mode data. One such respiratory correlated approach used a point source of 18F-FDG attached to the patient's skin to track respiratory motion [50]. Identifying the frames in which the point source fell within an operator defined region of interest (ROI) allowed PET images corresponding to different points within the respiratory cycle to be created. It was demonstrated that this technique produced similar results to gating. Another approach which uses time activity curves generated from a ROI drawn over a moving object in the image to recover the breathing frequency is currently undergoing clinical validation [49]. The advantage of these techniques is that the data may be retrospectively reconstructed for any breathing phase or amplitude.
Irrespective of the gating methodology implemented, the emission data acquired in each of the temporally gated frames is reasonably free of respiration-produced inaccuracies. However, the resulting individual frame images are of reduced resolution, as well as overall quality, as they contain only a fraction of the counts available throughout a PET acquisition [51]. Some groups have attempted to deal with this problem by acquiring gated data in 3D mode [45]. The need, therefore, exists for the development of correction methodologies making use of the gated datasets, in order to obtain respiration free PET images using all available data throughout a standard respiration average PET acquisition. This approach will also remove the need currently existing in terms of significantly increasing the time (over a factor of 3) of gated PET acquisitions in order to compensate for the presence of reduced statistics in the final reconstructed images. Very limited work is currently available in this domain. First, an emission driven solution through the combination of respiratory synchronised emission datasets and an iterative reconstruction algorithm can be envisaged, in a similar fashion to the methodology that has been previously suggested for SPECT cardiac imaging applications [52, 53]. The second option is based on a realignment methodology to "bring" all of the respiratory synchronised PET datasets to a particular phase in the respiratory cycle. This methodology is potentially applicable to both image and raw data domains, deriving the transformation parameters from the corresponding respiratory motion synchronised CT frames [48, 54].
|
The impact of PET imaging on patient outcomes |
|---|
|
TopAbstractIntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
Nioutsikou et al [56] have published an evaluation of the impact of including lung functional data into the radiotherapy planning process with a view to optimizing fields to exclude functional lung volume wherever possible. Whilst this is not directly related to PET volume definitions, it provides a methodology which could be applied to enable normal tissue toxicity to be evaluated when treatment plans are modified, thus building upon well characterized tissue toxicity data. Miften et al [57] had previously published a similar study using lung functional data to modify treatment plans. Van Der Wel et al [60] undertook a modelling study of the toxicity effects of the introduction of PET into stage N2-N3M0 NSCLC treatments. The group defined a therapeutic ratio and again demonstrated that in a group of patients tumour doses could be escalated at no detriment to normal tissues. The model was based rigidly upon ICRU guidelines for treatment volume definitions [28] with a rigid protocol for the introduction of PET data. All volumes used were based on the CT structures, with the CT defined structure removed if the PET scan was negative and CT structures added if the PET scan was deemed to show a positive uptake. No attempt was made to define volumes based on the segmentation of the PET data.
It would appear essential that these methodologies are used to inform the construction of clinical trials aimed at rigorously assessing patient outcomes and the associated cost benefits of such developments.
|
Conclusions and future work |
|---|
|
TopAbstractIntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
The process of integrated data acquisition requires the skills of therapy radiographers to ensure continuity through the treatment pathway. Experience has shown that if the PET/CT data are to be used directly in treatment planning, then the therapy radiographers will acquire a radiation dose from the close proximity work with the radioactive patient. This dose is not insignificant and is not readily avoided, unless the PET data are not acquired in the exact treatment position and introduced into the treatment planning process through software image fusion.
The utilization of the PET functional data still remains a challenge. The unique nature of the PET/CT dataset still presents many opportunities to use the two data types to optimize the definition of tumour volumes as the first stage of the radiotherapy treatment plan. At present, most centres have relied on the assessment of volumes from PET images based upon simple or adaptive thresholding methods validated with reference to spherical objects. These methods are relatively crude and do not take account of respiratory and cardiac motion, which effectively modify volumes in the lung and abdomen. There is also the potential to utilize the CT data to enhance the PET image to assist in automatic segmentation.
Physiological motion may have a significant impact upon the reconstructed PET image as the misalignments between the PET and CT datasets will result in an incorrect attenuation correction being applied which, in the chest in particular, will lead to incorrect tracer distributions [59]. This limitation has led to the active investigation of the use of respiratory gating in the acquisition of both the CT and PET datasets. The successful implementation of this technology has the potential to fully map the motion of tissue within the treatment fields and to provide aligned, correctly reconstructed, PET functional data. Commercial systems are already available, although their implementation is not straightforward. It should be noted that respiratory gated CT will potentially lead to a significantly increased CT dose to the patient. The gating of the PET acquisition will lead to a reduction in the signal to noise ratio in the images spanning the respiratory cycle as it will not be possible to increase the injected dose or increase the scanning times to compensate for the reduction in image acquisition time. It has been shown that this will lead to increased errors in volume determinations, which may outweigh the gains of better localization through gating. It must be remembered that the aim is to segment the lung based upon periods of equal volume and not to segment the respiratory cycle into a fixed number of time bins. Commercial systems are currently based on time or phase segmentation as the trigger systems do not produce lung volume related signals. Systems based upon chest expansion, such as a chest strain gauge, have the potential to provide a signal related to lung volume and to permit segmentation based upon phases of equal volume. For all of these methods the process must be capable of controlling the delivery of the radiation therapy to best utilize this input data.
This phase of development is aimed at the more accurate definition of treatment fields linked to the ability to conform the treatment fields to these defined volumes using IMRT and associated techniques. The introduction of functional maps into the process is, in itself, a confounding variable. An accurate understanding of the meaning of the image is vital to the definition of protocols which permit the routine and reproducible use of the data. Limited sensitivity and specificity for the PET signal will mean that some PET tracers will have very limited use in radiotherapy planning for some tumour types in some organ systems. It must also be remembered that the accuracy of PET interpretation, as with CT data, will be observer dependent and will almost certainly require the formation of teams of specialists to adequately and reliably incorporate PET data into the planning process.
The integration of functional data still presents the potential to better target external beam therapies. Dose escalation to hypermetabolic or radio-resistant areas may also be guided by specific functional markers. Evidence of the effectiveness of these methods in patients is still limited and at least one publication suggests that extreme caution is required if outcomes are not to be detrimental to the patient. There remains an urgent need to develop well validated protocols for the interpretation and use of PET functional images in the specification of radiotherapy treatment plans. These should be tested in well constructed clinical trials which themselves have been assessed by the modelling of radiation toxicity effects associated with these technology developments.
Received for publication July 11, 2005. Revision received October 26, 2005. Accepted for publication January 25, 2006.
|
References |
|---|
|
TopAbstractIntroductionThe PET image: its...Target volume delineation for...Respiratory gating for PET/CT...The impact of PET...Conclusions and future workReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  BJR review of the year - 2006 Br. J. Radiol., March 1, 2007; 80(951): 147 - 151. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| BJR | DMFR | IMAGING | ALL BIR JOURNALS |
