Image guidance for precise conformal radiotherapy

Thomas Rockwell Mackie     IJROBP 2003;56:89

Image-guided conformal therapy can mean many things. The use of images in radiation therapy is not new. Treatment planning has been using film, fluoroscopic, and even CT images for decades. Image-guided precision conformal therapy here means the continuous quality improvement using images to better achieve the goals of radiation therapy, which is to deliver a tumoricidal dose of radiation without unduly harming normal tissue. Images have evolved from pictures of the patient to become computer-based representations of the patient. Conformal radiotherapy to some means the use of blocks in radiation delivery. In fact, traditional blocking is intensity modulation using only two intensity levels. Intensity-modulated radiation therapy is rapidly approaching perfection in conformal delivery to the tumor and conformal avoidance of sensitive structure, the precision limited only by the fidelity and acuity of the image representation of the patient before and during the course of treatment.

The fear of collateral damage to neighboring tissue often limits the dose that can be applied to the tumor. With the advent of three-dimensional (3D) imaging modalities, the location of many anatomic structures within the body, as well as their function, can be mapped with reasonable precision. With 3D conformal radiotherapy and intensity-modulated radiotherapy (IMRT), it is possible to optimize the delivery of radiation to complex target volumes. Most IMRT systems achieve this objective by using multiple beam directions with the use of several small segmented or modulated fields. IMRT’s introduction of numerous small radiation fields inherently increases delivery inaccuracies. As a consequence, use of IMRT without precise localization of the tumor and sensitive structures, at both the planning and delivery stages, and the absence of continuous verification represent the most significant challenges to the implementation of IMRT in routine clinical use.

In the past few decades, 3D imaging modalities such as CT, MRI, and positron emission tomography (PET) have allowed reasonably precise localization of the gross tumor volume (GTV) for many sites and for most sensitive structures. Location of microscopic extension of the tumor around the GTV remains uncertain. The concept of clinical target volume (CTV) surrounding the GTV with a presumed margin for microscopic invasion is a best-guess scenario to overcome this deficit. Because the patient is not set up identically from day to day, and the position and shape of organs may shift and deform, a volume larger than the CTV called the planning target volume (PTV) has to be treated. The margin of the PTV around the CTV has to account for setup uncertainty and organ motion. Three-dimensional conformal radiotherapy treatment planning, which is one example of an image-guided system, has to plan for these uncertainties.

Future 3D image–guided radiotherapy systems must do more than plan for uncertainties; they must endeavor to eliminate their impact. There are two general strategies that can potentially achieve this objective:

1.Reduce the uncertainty of microscopic extension (i.e., where information supports it, CTV → GTV).
2.Reduce the uncertainty in the setup and identify and reduce organ motion (i.e., when possible, PTV → CTV).

Both of these strategies taken together will reduce the margins around the GTV. In effect, margins blur the radius of curvature of GTV contours to a value equal to the margin width. Today’s large margins moderate the demand for high-resolution planning, delivery, and verification systems. As margins shrink, there will be a need for tighter resolution requirements of image guidance systems.

Recent evidence indicates that some sites may respond better to higher doses per fraction. The α/β ratio for prostate carcinoma may be relatively small  with estimates as low as 1.5 instead of the 8 to 10 typical of tumors. In lung cancer, there is evidence that the dose per fraction should be increased, and the overall time should be shortened. In some cases, a higher dose per fraction can be safely delivered only if the dose per fraction to normal tissues is not significantly increased. This will increase the importance of image guidance. Because a higher dose per fraction will reduce the number of fractions delivered, the increased effort of image guidance is justifiable.

Image-guided IMRT can potentially achieve improved radiotherapy by developing appropriate planning, delivery, and patient-specific verification systems and processes with an adaptive flexibility. One such system, helical tomotherapy, is described as a new approach to conformal adaptive image-guided IMRT.

Improving the definition of the clinical target volume

Reducing the uncertainty of microscopic extension of the tumor is a difficult task. Even today, the CTV margin for some sites, such as prostate and breast, may perhaps be too generous, whereas for other sites, such as glioblastoma multiforme, it may be too small. What is needed is a three-pronged research attack on the problem:

1.High sensitivity and potentially high-specificity imaging systems, such as PET, correlated with image-registered pathologic specimens must be used to help delineate disease extent from normal tissue at the tumor periphery, so that knowledge of the probability of spread can be determined.
 
2.Outcome analysis has to be precisely quantified, so that it can be determined whether patients are failing at the margin or within the tumor. This will be a tremendous complement to dose escalation trials, because even extremely high doses will not be sufficient to eradicate with high probability tumors that are routinely missed.
 
3.Begin treating patients with a conformal avoidance methodology in addition to conformal radiotherapy; this allows for the use of modestly large CTV margins while still sparing critical normal structures.

Imaging systems continue to improve with respect to sensitivity and specificity of diagnosis. Radiation therapy treatment planning requires sensitivity and specificity on a voxel-by-voxel basis. CT alone, or fused with other imaging modalities, is still the gold standard for delineation of most tissue structures. Improvement in CT contrast agents has allowed much better definition of tumor boundaries. CT simulation with 3D treatment planning is now conventional practice in most radiotherapy centers, and the need for conventional simulators for some treatment sites has been called into question. The sale of dedicated CT simulators is now larger than the sale of conventional radiotherapy simulators. Even with dedicated 3D imaging systems common in the clinic and a generation of radiation oncologists trained in their use, there is considerable variation of the clinical target volume for many sites.

The ability to differentiate cancer from normal and noncancerous pathologies has improved greatly with the advent of MRI. MRI is now the definitive way by which most central nervous system malignancies are delineated. Magnetic resonance spectroscopy (MRS) is being used to pinpoint the active disease site within the prostate, in effect, identifying the GTV within the CTV. Currently, MRS voxel sizes are rather coarse, but better signal processing and higher field strength magnets will enable the voxel sizes to shrink to those comparable to dose computation resolution. Selective boosting of the active disease will require excellent tumor and target localization. It is likely that MRS will find uses in many tumor sites.

The role of PET will be greatly enhanced by the new generation of PET/CT scanners from which fully correlated PET and CT images emerge. PET/CT simulators will augment or even replace CT simulators during this decade. PET images will be able to delineate the extent of the GTV and the CTV with much more precision. PET will begin to enable a probabilistic formulation for the GTV and CTV whereby the GTV is that region where the probability of occupancy by tumor approaches 100%, and the CTV is at the boundary where that probability approaches zero. The PET successors to Prostascint scans might identify minimally involved lymph nodes in prostate cancer. Radiopharmaceuticals to complement 18FDG, a good marker of proliferative metabolic activity in some cases, will emerge. Labeling of proliferation, hypoxia, perfusion, even the status of the P53 gene in the tumor, will be possible.

Although better diagnostic imaging will aid in improving GTV definition, there will always be a residual uncertainty at the tumor boundary. Individual patients’ pathologic specimens can reveal quantitative information regarding the extent of microscopic disease. Prostate needle biopsies that document the path of the needles on 3D ultrasound images would be extremely useful. Detailed studies with current technology will be labor intensive and expensive. Computer-aided diagnostic systems have recently been approved for mammography. Similarly, intelligent systems that automate the screening of pathologic samples and correlate this information with imaging studies need to be developed to gather information on nodal disease spread on individual patients and populations of patients, so that rational margins around the GTVs can be established. Image guidance has to be carried into the microscopic realm. There is a great opportunity for the molecular medicine revolution to fight its first battles in a better-controlled pathology arena.

Radiation therapy has not routinely gathered highly accurate information on distributions of failure in and near the treatment volume, and that information is needed even more now. As CTV margins around disease sites are reduced and doses are escalated, there must be careful monitoring to determine whether the patterns of failure are changing. Smaller margins will result in more motion blurring of the dose inside the target volume. There are plenty of data on failure from autopsy series; the problem is that the fine structure of local failure is blurred by overgrowth of the recurrence. Without image-based follow-up correlated to the treatment plan to provide in vivo recurrence detection, failures could masquerade as a consequence of inadequate dose instead of inadequate margin. Shrinking the margins for the sake of shrinking the margins when there is no critical sensitive tissue to spare might lead to a decreased therapeutic ratio.

Conformal avoidance is the complement to conformal therapy. The strategy of conformal avoidance is to treat the tumor with generous treatment volumes, but to carve out dose-limiting regions around sensitive structures where there is no possibility of tumor extensions. Conformal therapy should be used when the GTV is well delineated and the CTV extension is small. Conformal avoidance should be used when the GTV is not well delineated or the CTV is large. This strategy was not possible before the advent of IMRT, because it was extremely difficult to produce arbitrarily shaped treatment volumes with standard uniform field irradiation. Examples of conformal avoidance include sparing uninvolved parotid glands, auditory apparatus, mucosa, and larynx in head-and-neck radiotherapy. In advanced prostate, the paraprostatic nodes are the ones at highest risk. A treatment envelope could cover the nodes but avoid the rectum and the lower bladder.

Conformal avoidance radiotherapy requires image-guided delivery. Carving out regions around sensitive structures will result in placing high gradients near the sensitive structures. To have confidence in this strategy, the second major objective for precise delivery of radiotherapy has to be met: the reduction of setup uncertainty and organ motion and confirmation of accurate dose delivery.

Reduction of setup uncertainty and organ motion

Just as the margin extension around the GTV to form the CTV should be based on the notion of probability of inclusion, so should the PTV extension of the CTV. Where possible, setup uncertainty and organ motion should be reduced, but where not possible, treatment planning systems should include error analysis based on positional variance to be included, so that most likely dose distributions and expected dose–volume histograms can be determined for the target volumes.

Image-guided delivery of radiation therapy is required for precise conformal therapy or conformal avoidance therapy for many treatment sites. There are examples of radiation therapy today for which setup uncertainty or organ motion is not an issue. In stereotactic radiosurgery applied to the brain, the use of imaging to localize the disease with pinpoint precision is justified, because the brain is held relatively fixed within the cranium, and the cranium is fixed to rigid external devices or implanted markers. When it can be guaranteed that the representation of the patient at the time of delivery is the same as at the time of imaging, and the tumor volume is small, one or a few high doses of radiation can be delivered to the tumor. However, even in stereotactic radiosurgery, imaging can be used to verify that the fiducial frame position did not become altered between planning and delivery.

In head-and-neck radiotherapy, immobilization and optical-guidance systems  can greatly reduce the need for image guidance, but may not eliminate it completely. Optical guidance is accomplished through detection of four passive markers that are attached to a custom bite plate that links to the maxillary dentition of the patient to form a rigid system. A camera system mounted to the ceiling of the linear accelerator vault and interfaced with a computer tracks the translations and rotation of the patient’s head. Optically guided radiotherapy may enhance normal-tissue sparing, provide a high degree of conformality, and improve dose homogeneity characteristics in head-and-neck patients.

Stereotactic radiosurgery can be applied to the body as long as the patient can be localized with precision at the time of treatment. Aggressive fractionation schedules similar to those used cranially can, in principle, be applied to extracranial sites as long as the treatment volume is small, and normal structures are avoided. Stereotactic body frames attempt to limit the setup uncertainty and the range of intrafraction motion so that the PTV can remain small. Image guidance at the time of treatment will increase the certainty that the tumor is being covered and sensitive normal tissue avoided.

It has been well established that the prostate moves considerably between fractions, mainly because of differential filling of the rectum with fecal matter or gas. The direction of motion of the bulk of the prostate is mainly in the anterior-posterior direction with accompanying shape distortions. Conventionally, this has been accounted for with generous margins, which results in considerable dose to the rectum and bladder. Immobilization of the prostate using a rectal balloon is another approach to improving the setup and enables exploitation of smaller margins.

It is routine to use image guidance in brachytherapy. In brachytherapy prostate treatments using radioactive seeds, there is little impact of organ motion, because the seeds are moving along with the prostate. This technique is able to give much higher initial brachytherapy doses to the prostate than is safe for conventional external beam radiotherapy. However, a higher initial dose may be needed in prostate seed brachytherapy, because edema that subsequently may result from the procedure lowers the effective dose delivered. If external beam radiotherapy could reduce the effects of setup uncertainty and organ motion, then biologic doses approaching those actually given in prostate brachytherapy could be delivered to the tumor, because significantly less radiation would be given to normal structures.

The imaging system typically employed on a linear accelerator today is a portal image, which is a type of planar radiography system. In general, the portal image is acquired with radiographic film, but increasingly, electronic portal imaging detector (EPID) systems are being used to image the beam exiting the patient. EPIDs are an improvement over film, because they are more efficient and offer more latitude for image enhancement. EPIDs have been used to automatically correct the patient setup, for example, using a “tilt and roll” couch. Many investigators have shown that it is possible to collect the “exit dose” from the EPID signal. In principle, if processed fast enough, this dosimetry information could be compared to the exit dose calculated from the treatment plan and used as the basis to halt the treatment.

Even the best portal images will be inferior to a typical kilovoltage planar radiograph, because there is far less tissue contrast in a megavoltage image. Planar radiographs are difficult to interpret, because they are not in reference to three-dimensional CT image sets, and out-of-plane rotations of the patient are not as readily evident as translations. Three-dimensional conformal radiotherapy planning systems can augment the poor image quality of planar radiographs and improve their comparability with CT by the construction of digitally reconstructed radiographs from CT  or even magnetic resonance image sets in the same treatment setup and beam position as the radiographs. However, when mismatches occur between digitally reconstructed radiographs and planar radiographs, the reasons are usually not readily obvious or easily determined. Radiopaque invasive markers can be placed in certain structures such as the prostate to assist in its localization with EPIDs , but this is a highly invasive procedure and has limited applicability.

For some time, megavoltage image detectors have been used to acquire a set of projections around the patient so that megavoltage CT (MVCT) images are reconstructed. Cone-beam MVCTs could potentially be obtained from conventional linacs using portal imaging systems, although the dose needed to produce the same image quality would be higher. There are a number of advantages of MVCT over planar megavoltage portal radiographs:

1.MVCTs are fully three-dimensional.
2.MVCTs have better soft-tissue contrast than planar radiographs.
3.MVCTs are easier to compare with planning CT images.

In addition to using the treatment beam to produce an MVCT scan, it is possible to acquire a kilovoltage CT (kVCT) scan at the time of treatment. Several approaches are being investigated. It is possible to put a conventional CT scanner in the treatment room. The CT scanner may be driven on rails over the treatment couch. Alternatively, a couch may be designed to transport the patient either through the bore of the CT gantry for imaging or beneath the linear accelerator for treatment. The tissue contrast visible at a given resolution and given dose for a conventional kVCT is superior to that for an MVCT. It is possible to put a kilovoltage X-ray tube and a detector array on board the linac to acquire CT scans. Because the gantry rotation of a linac is much slower than a CT ring gantry, flat-panel detectors are more practical from the perspective of throughput, so that a whole volume may be acquired with one or a few rotations. These so-called cone-beam CT scanner systems do not have as favorable imaging properties as conventional kVCT scanners, because there is more scatter arriving at the detector that degrades the image contrast. The image contrast can be restored by an increase in the dose to the patient or an increase in the slice thickness.

Even with image guidance immediately before or during treatment, intrafraction organ motion may limit the accuracy of delivery. Organ motion is potentially most severe in the thorax where respiratory excursions move parts of the lung and liver appreciably, whereas other thoracic structures are less mobile. Even the prostate may move during breathing. Patients in the prone position seem to have a larger range of prostate motion due to breathing compared to patients in the supine position.

One approach to overcoming respiration-induced motion is gated delivery, which applies the radiation only when the organ is in a preselected position or state. Gated delivery is easiest for conventional uniform field delivery, although it may also be possible with simpler forms of IMRT, such as static IMRT techniques (so-called “step-and-shoot” methods). If the patient can tolerate breath holding for a reasonable period, other methods, such as maximum breath-hold delivery, can be pursued, and additional advantages then accrue. During normal breathing, the lung is in the maximum expiration the longest fraction of the time; however, when asked to hold their breath, people typically deeply inspire. In addition, during maximum inspiration, the lung is fully inflated, so that normal alveolar tissue is maximally displaced from the high-dose regions. If the patients’ health status allows them to perform breath holding, but they are uncooperative, it is possible to use active breathing control to control the movement of the lung. To date, delivery of IMRT using gating or breath holding has necessitated that field delivery be halted after only a fraction of the volume has been treated, to allow the patient to breathe. This “segmental delivery” approach may lead to hot or cold regions in the tumor.

Another potential solution to the problem of organ motion under development is “tumor chasing” or “tumor following” using dynamic aperture tracking. The displacement of the lung in a breath-hold cycle can be as large as 2 to 3 cm. It is greatest at the base and least at the apex. There is hysteresis (discordance between expiratory and inspiratory excursions) in the lung motion, because the track of the lung during inspiration is not the same as that during expiration. One technique favored at the current time is to put one or more fiducial markers in the tumor and use multiple stationary ceiling-mounted X-ray tubes and floor-mounted detectors to monitor the movement of the fiducial . Real-time fluoroscopy and active control of the couch or the beam is then used to keep the fiducial markers at the same location relative to the treatment beam. This can be accomplished either by moving the patient or moving the beam. Moving the patient at the required rates could be potentially unsettling for many patients. It is also possible to move conventional MLC leaves by the 1 to 2 cm per second required in the superior-inferior direction, but it is difficult to move conventional MLCs simultaneously in two directions.

Perhaps the ultimate in radiation delivery to the lung is to use a four-dimensional model of the lung as the fundamental representation; this obtains traditional geometric information in three dimensions at multiple time points. This can be acquired using a fast helical CT scanner or multiple fast MRI pulse sequences. Acquiring the planning images and delivering the treatment would require a cooperative patient or the use of active breathing control. Treatment planning would involve dose calculation to the lung in a few phases of the breathing cycle, from inspiration through expiration and back. During treatment, an image guidance or tracking system would continuously determine the phase of breathing. The patient would be coached to reproduce as closely as possible the breathing pattern for which he was planned. The appropriate delivery for the phase would be chosen in real time from the planned set of delivery patterns corresponding to each breathing phase. This system would require accurate verification of the dose distribution correlated to the actual patient representation to ensure that the treatment was correctly delivered. The ability to obtain tomographic information during delivery would enable real-time tracking without the need for implanted markers.

Helical tomotherapy: a dedicated image-guided IMRT system

Helical tomotherapy represents the fusion of a linac with a helical CT scanner. With this system, a fan beam may be used to acquire an MVCT of the patient just before treatment and potentially even during treatment. During treatment, a dedicated binary MLC is used to modulate the same fan beam to provide rotational IMRT. The beam rotation is synchronized with continuous longitudinal movement of the couch through the bore of the gantry, forming a helical beam pattern from the patient’s point of view. The set of binary collimator leaves rapidly transitions between open (leaf retracted) and closed (leaf blocking) states. When operating as an MVCT system, the leaves are fully retracted to the open state.

Putting a linear accelerator into a ring gantry means that noncoplanar treatments for sites in the body are not possible. There are times when noncoplanar beams are useful, especially when the tumor is surrounded by a normal tissue structure, and the noncoplanar distances through the normal tissue are less than or equal to those for the coplanar directions. This is most often applicable to the brain, but it may also be true for the lung in some circumstances. With tomotherapy, it still may be possible to do limited noncoplanar delivery to the head using the natural capability of the head to tilt, as was once done before CT gantries had the ability to tilt.

There are a number of verification processes that are possible with a CT scanner on board. Immediately after the MVCT scan, the scan can be fused with a planning CT scan to determine if the patient is set up correctly. The fusion process must be done quickly enough to be useful, but carefully enough to be accurate. A combination of automated and manual fusion techniques is essential, because the operator must be able to override the automated procedure, if necessary. The automated fusion routine was able to find the correct translations and rotation in only a few seconds per slice to an accuracy of less than 1 mm or 1 degree. If the patient is not set up correctly, the translation and rotation information obtained from the fusion can be used to correct the setup. The setup correction involving rotation and translation (so-called rigid transformations) can be implemented either by moving the patient or, in principle, by modifying the IMRT delivery to account for the patient’s actual geometric offset. More general affine transformations, which include scaling and shearing in addition to translations and rotations, could be achieved with modifications to the leaf delivery pattern. Even more complex transformations are possible, although computer speed will likely be a limitation for the foreseeable future.

The capabilities of having a CT image at the time of treatment will not be fulfilled without image analysis software. Three-dimensional treatment planning analysis requires segmentation of relevant structures. The most time-consuming aspect of 3D treatment planning is the segmentation of the planning image. It is unreasonable to expect physicians to repeat their segmentation tasks routinely. Automated tools to segment the relevant volumes on treatment CT images are more practical in this application, because the planning CT will have been segmented and can be used as a template for autosegmentation. Computers cannot be expected to automatically render contour volumes in an unsupervised manner. Physicians, or those well trained in anatomy, will be required to approve the automated procedures and to correct mistakes. Similar to automated processing of scanned paper documents into text, the utility of these systems will depend on their error rate and on how easy it is to correct them.

Acquiring a CT scan immediately before treatment puts less emphasis on patient localization devices, such as stereotactic coordinate frames, and more emphasis on patient immobilization. The CT representation at the time of treatment forms its own localization system. What is important is to guarantee that the patient does not move during the treatment. The CT detector system on the helical tomotherapy unit is operating during the treatment. Delivery verification can compare the detector signal during treatment with the expected signal. A sudden deviation in the signal from that expected could be used as the basis for treatment termination.

Operating the CT detectors during treatment can also be used to reconstruct the dose delivered to the patient. Dose reconstruction is the determination of the dose delivered to the patient on the basis of detectors that determine the exit dose. Dose reconstruction requires an accurate CT representation of the patient at the time of treatment. The exit dose map and the CT representation just acquired are used to determine the energy fluence distribution incident on the patient. The energy fluence distribution and the CT representation can be used to compute the dose distribution in the patient. This reconstructed dose distribution represents the dose the patient actually received, and it may be superimposed on the CT representation just obtained. Therefore, the dose reconstruction can be directly compared with the planned dose distribution superimposed on a planning CT to provide unprecedented treatment verification.

Consistent access to an accurate delivered dose distribution opens up the possibility of correcting or adapting future treatment delivery to take into account the deficiencies of previous dose distributions. One complicating factor is that the patient representation in the original planning CT might not be the same as the representation at the time of treatment. How is it possible to determine the difference in dose to an organ when the shape of the organ may be different between the two representations? One approach to this is to deformably register the two patient representations together. This could be achieved by determining reliable features in both image sets and using these features as the boundary conditions for a finite element algorithm for deformation. Such features can be contours drawn by the physician or points, lines, or surfaces detected automatically. As with automated contouring, physicians or trained anatomists will have to approve the results of deformable registration. The process of applying feedback to the image-guided radiotherapy processes is called adaptive radiotherapy. The best way in which this feedback can be applied is under active research.  Is there an optimal way to estimate the corrections when the verification results are uncertain? Methods from statistical signal processing can be applied to find, in a statistical sense, the most likely position of an organ based on an uncertain setup verification measurement. Basic control theory predicts that corrections should be damped. Potentially, the correction could be effected over several treatments. How should delivery be calculated? Should the original prescription be modified? These questions will be answered as experience is gained.  As computers become faster, the adaptive radiotherapy processes will become more powerful, enabling full reoptimization of dose distributions before each treatment and adapting to organ motion while it occurs. Helical tomotherapy provides unique solutions to the problems associated with delivering IMRT to sites in the thorax. We have recently demonstrated that it is feasible with helical tomotherapy to deliver a nearly uniform but smaller than adequate dose to an entire lung tumor during a single 10-s breath hold. This is achieved with a loose helical delivery technique. The full treatment dose is delivered using multiple helical trajectories, one per breath-hold, with the starting point of the helical trajectory displaced rotationally between each delivery. In this way, there would be no need to junction the beam boundaries within the target volume.

There is yet another way in which helical tomotherapy could account for thoracic motion. On the helical tomotherapy unit, the jaws defining the field width could track in the superior-inferior direction corresponding to the majority of the respiratory motion and still account for residual lateral movement using the fast-moving binary collimator. This would allow tumor motion to be tracked without oscillatory movements of the patient.

Discussion

Helical tomotherapy is a linac on a CT ring gantry. As discussed earlier, it is also feasible to put a CT scanner onto a conventional C-arm linac gantry. The field apertures of conventional linacs support cone-beam CT using electronic portal imaging systems. With one rotation, many CT slices can be obtained at once. At megavoltage energies, the quantum efficiency of portal image systems is only a few percent, and so currently the dose delivered to the patient would be too high for daily use. In the future, helical tomotherapy units may employ multirow detector systems just as helical CT systems do now. It is likely, in the future, that as communications and data processing speed increases, the number of CT rows will increase on CT scanners until they have as many rows as a portal imaging system. Meanwhile, cone-beam CT on conventional accelerators will face the difficulty of correcting for gantry sag and ultimately slow rotation speeds, because the C-arm gantries rotate so slowly. The only way to safely increase the speed of rotation is to enclose the gantry with stationary ring cowlings, thereby looking like a tomotherapy unit and losing the ability to deliver noncoplanar treatments. Both tomotherapy and conventional units may be equipped with kilovoltage sources in the future. Therefore, in the long term, there will be technical pressure for tomotherapy and conventional units to develop toward the same solutions.

Helical tomotherapy technology is new. As such, it is largely untested clinically. At the time of writing, the first clinical unit was just beginning animal and human clinical trials. There will undoubtedly be unforeseen teething problems with this new technology, but the greatest hurdles to the acceptance of this and all image-guided technologies are not technical but psychological in that they challenge the established paradigms of radiation therapy, and some degree of retraining will be required to master them. However, if tomotherapy and other image-guided radiotherapy technologies demonstrate that tumors and sensitive structures can be reliably localized, these newer paradigms will become accepted, and the difficulties involved in retraining will be gladly surmounted.

Conclusions

The goal of radiation therapy is to eradicate the tumor bed while sparing healthy tissue. The first aim must be to delineate tumor from healthy tissue. Image-guided radiation therapy will be able to reduce the uncertainty of microscopic extension of disease. Advanced imaging systems correlated with image-registered pathologic specimens will be able to delineate disease extent from normal tissue at the tumor periphery. Outcome analysis to determine whether patients are failing at the margin or within the tumor must be routinely carried out. When it is not possible to determine the clinical target margin with reasonable certainty, the margins must remain generous, and conformal avoidance methodology must be used to spare critical normal structures.

As important as defining the clinical target volume is, the need to guarantee that it is indeed being treated is as important. The patient must be able to be set up for treatment in as reproducible a manner as possible. Image guidance using a variety of systems, including portal images, ultrasound devices, and CT scanners, has been implemented at the time of treatment. Some image-guided methods, portal images for instance, are more amenable for use with rigid structures such as encountered in the sinus, whereas other methods, such as ultrasound or CT scanners, are able to account for nonrigid setup variations. Several strategies for preventing organ motion from degrading the precision with which radiotherapy can be delivered are under investigation. They include monitoring the position of implanted fiducial markers using both portal imaging and fixed X-ray systems.

Helical tomotherapy has been designed for image-guided intensity-modulated radiation therapy. An on-board megavoltage CT scanner enables verification CT scans to be acquired before treatment. The CT scan set can be automatically fused with a planning CT to determine the 3D shape and position of the target volume before radiotherapy. This information can be used to adjust the patient setup, if necessary.

A CT scan at the time of treatment delivery can also be used as the basis for reconstructing the dose received by the patient. Dose reconstruction will allow the dose just delivered to be superimposed on the CT scan just acquired and compared with the planned dose distribution superimposed on a planning CT. If the anatomy has been altered by nonrigid deformation, deformable registration can be used to map the delivered dose distribution back onto the planned dose distribution. Dose reconstruction with or without deformable registration will allow the dose delivery to be modified on subsequent treatments. This set of verification processes that strives to deliver correctly throughout the whole course of radiation therapy is called adaptive radiotherapy. Table 1 is a summary of methodologies to effect precision radiation therapy.