Intensity Modulating and Other Radiation Therapy Devices for Dose Painting

James M. Galvin, Wilfried De NeveJournal of Clinical Oncology, Vol 25, No 8 (March 10), 2007: pp. 924-930

The introduction of intensity-modulated radiation therapy (IMRT) in the early 1990s created the possibility of generating dramatically improved dose distributions that could be tailored to fit a complex geometric arrangement of targets that push against or even surround healthy critical structures. IMRT is a new treatment paradigm that goes beyond the capabilities of the earlier technology called three-dimensional radiation therapy (3DCRT). IMRT took the older approach of using fields that conformed to the silhouette of the target to deliver a relatively homogeneous intensity of radiation and separated the conformal fields into many subfields so that intensity could be varied to better control the final dose distribution. This technique makes it possible to generate radiation dose clouds that have indentations in their surface. Initially, this technology was mainly used to avoid and thus control the dose delivered to critical structures so that they are not seriously damaged in the process of irradiating nearby targets to an appropriately high dose. Avoidance of critical structures allowed homogeneous dose escalation that led to improved local control for small tumors. However, the normal tissue component of large tumors often prohibits homogeneous dose escalation. A newer concept of dose-painting IMRT is aimed at exploiting inhomogeneous dose distributions adapted to tumor heterogeneity. Tumor regions of increased radiation resistance receive escalated dose levels, whereas radiation-sensitive regions receive conventional or even de-escalated dose levels. Dose painting relies on biologic imaging such as positron emission tomography, functional magnetic resonance imaging, and magnetic resonance spectroscopy. This review will describe the competing techologies for dose painting with an emphasis on their commonalities.

Three-dimensional radiation therapy (3DCRT) was considered radically new when it was first introduced because it integrated computed tomography (CT) imaging directly into the treatment planning process. Transferring planning CT data sets to the treatment planning computer provided a convenient method for shaping treatment fields on the basis of precise knowledge of the patient's three-dimensional internal geometry. Given this improved information, 3DCRT beams are fashioned with tight margins to conform to the projected outline of the target or combination of targets. However, a major limitation of 3DCRT is its inability to account for indentations in the target where critical structures invaginate the surface of this volume.

When intensity-modulated radiation therapy (IMRT) was introduced in the early 1990s, it represented an important advancement because it addressed this shortcoming of 3DCRT. That is, it provided a mechanism for painting or sculpting dose distributions so that their 3D shape agreed with the shape of the targets even in the regions that are hidden from view when a series of two-dimensional "beam's-eye view" silhouettes are drawn around the target from different directions.

Instead of restricting the search for useful beams to the ones that conformed to the outline of the target or targets, IMRT took the 3DCRT conformal beams and divided them into many smaller subfields to create the ability to modulate intensity across these apertures. This change made it possible to partially shield any part of the target volume where a critical structure is hidden from view because it is surrounded by target tissue. For one projection, this partial-shielding technique will cause a shadow of low dose through the surrounding parts of the target. However, by repeating this process for a number of different projections, the shadow is spread across the target region and diminished in its effect. The end result is that the dose distribution for the target or targets is relatively homogeneous, whereas the dose lines are bent in and around the intruding critical structure. This describes the rather simple idea of how IMRT works, but describing a number of other aspects of the total process is necessary to provide a complete picture of how this technology was successfully implemented a number of years ago. Additionally, during the intervening period since the concept of IMRT was first introduced, researchers have devised a number of variations mainly in the area of dose delivery. These variations will also be discussed.

It is important to point out that some features of IMRT are not particularly new. However, to achieve the superb dose distributions with rapid dose fall-off that are characteristic of IMRT, these features have become an essential part of the total process. One example of the parts of the process that are not particularly new is the routine use of many different angles of approach for the various treatment beams that include noncoplanar field directions (ie, treatment beams with center lines that are out of the cross-sectional plane of the patient). Another example is the coupling of the delivery of these beams into a single button push.

The new features of IMRT that are not normally associated with previous treatment techniques like 3DCRT are (1) the inverse treatment planning process that works from a set of dose constraints for the target(s) and critical structures and (2) the use of a large number of treatment fields or subfields that fall within the conformal beam's-eye view of the target.

Thus, IMRT consists of two related steps: computerized treatment planning and dose delivery. Inverse planning algorithms can be developed with the assumption that conformal beams are divided into much smaller independent subunits, but there remains the question of how these smaller beams might be efficiently delivered. The next section discusses the inverse treatment planning part of IMRT, and explains how separating conformal fields into smaller beamlets allows the dose to be painted throughout a target of complex irregular shape, even when different regions of the target are to receive varying doses.

There is a building interest in integrating biologic information into IMRT planning for the purpose of targeting radiation resistant regions inside the tumor.The procedure, called dose painting, results in a conservative use of radiation dose. Dose painting strives to tailor the dose inside the tumor to deliver the exact amount of radiation needed for eradication and challenges the dogma of dose homogeneity to the target. The rationale for dose painting in healthy tissues is equally clear. Unwanted but unavoidable dose should be painted for optimal preservation of function. Sparing of functionally critical parts of organs at risk is prioritized in the process. The vision is to exploit functional imaging to predict a volumetric map of radiobiological factors in tumor and healthy tissue that is, in turn, translated to a map of desired dose and dose constraints. With dose painting, the opportunities to improve the local control versus toxicity ratios seem substantial. However, the appropriate trials to demonstrate the clinical value of dose painting have not been conducted. It must be emphasized that the threats are multiple and reside mainly in insufficient knowledge regarding (1) the biologic meaning of functional imaging and (2) the uncertainties in the transformation of radiobiological factors to desired dose and dose constraints. Preclinical research is needed to formulate stable hypotheses for clinical trials. Clinical trials require appropriate technology for planning and delivery of dose painting. This technology is being developed mainly in the field of IMRT, but also in some dedicated equipment for high-precision radiation therapy such as Gamma Knife (Elekta Instrument AB, Stockholm, Sweden) and CyberKnife (Accuray, Sunnyvale, CA).


To better understand how IMRT works, it is helpful to follow the initial development of this technology. At least one early attempt to design a planning algorithm that started from a simple summary of the clinician's ideal dose distribution and worked toward deriving the treatments beams that would give the desired result borrowed heavily from existing CT scanning technology. CT scanning irradiates the portion of the body that contains the tissues of interest with a large number of individual beams emanating from an x-ray source that rotates around the patient so that a large number of different angular orientations of these beams is obtained. The beams are defined by a bank of detectors that divide a fan beam of radiation emanating from a point source into separated parts. These separate detectors allow individual measurements of the transmission for all of these beams, and the computerized reconstruction algorithm determines a pattern of densities that is consistent with the numerous measured values. Early investigators posed the radiation therapy question somewhat differently. They looked at the problem as one of irradiating the target region within the patient's body with a very large number of small beams (eg, beams with a cross section on the order of 1 x 1 cm) and used algorithms similar to the ones developed for CT to find the beam weights that would achieve, or at least closely approximate, a desired dose distribution. The weights determine the amount of radiation for each beam, and these are related to that beam's intensity. Thus, because the weights of each small beam are adjusted to give the optimum final solution, the process was called intensity modulation.

In its current form, which includes biologic modeling, planning for dose painting starts with line drawings of target volumes and organs at risk. The drawing effort results in contours  based on anatomic  and includes biologic image information. Planning aims for the volumes enclosed by the contours are usually set as dose objectives for the target volumes and dose constraints to the organs at risk. For painting the target, two different methods have been proposed. One method, called dose painting by contours, prescribes a global dose objective inside the biologic image-based contours. For example, the dose to a particular target might be prescribed as a dose value that has to be reached for a percentage (eg, 95%) of the volume together with an acceptable dose range in that target.

The other method, called dose painting by voxel intensity, prescribes the dose inside the target for each voxel as the function of the signal intensity of that voxel in the biologic image.The latter method is sometimes called dose painting by numbers.Similarly, for organs at risk, functionally active regions can be contoured and spared by setting global dose constraints, or dose constraints can be set voxel by voxel based on the functional activity of each voxel.

The inverse treatment planning algorithm uses the 3D map of dosimetric goals and constraints, called dose objectives for planning, to determine a machine instruction file. Ideally, the machine instruction file allows the radiation equipment to deliver exactly the desired dose painting.

In modern oncology, a dose (provisional) prescription is usually written in a clinical protocol, which also specifies the patient cohort by means of patient-selection criteria. The term dose provisional prescription is the expression of the desired dose distribution in a specified patient cohort or in an individual patient for whom the planning process has not been started. IMRT planning is performed in the virtual world into which the individual patient is brought by imaging. On the basis of the patient-specific imaging, impossible, irrelevant, or conflicting dose objectives may exist in the dose provisional prescription. Inverse planning systems differ substantially in their ability to deal with such problematic dose objectives. The planner usually removes conflicting dose prescriptions by clinically appropriate priority ranking. For example, if a target volume overlaps an organ at risk and if the minimum dose objective in the target exceeds the maximum dose constraint in the organ at risk, the planner will assign priority either to the target or to the organ at risk. However, planners may choose to leave ambiguous dose objectives for planning as an optimization problem to explore the range of dose distributions that are physically possible to achieve. Other planners will remove problematic dose objectives as much as possible. The second approach will deliver more predictive plans, but at the expense of missing interesting dose distributions by reducing the domain of search. It is unknown which is the best approach.

In dose painting by numbers, the desired dose can almost never be achieved physically for each voxel. Achieving the closest possible approximation to the desired dose painting then becomes the goal of the inverse planning process.


During the same time period when inverse treatment planning algorithms were under development to make them robust enough for routine clinical application, researchers started to consider the problem of finding a practical method for irradiating a patient with a very large number of beams with individual weights that could be easily varied. One of the early approaches to solving this problem, again borrowed from CT scan technology, is adopting a fan-beam geometry and rotating this field arrangement around the patient to produce a large number of individual beams passing through a thin section of the body. As was the case for the CT units in use at that time, the patient was then moved slightly so that more beams were added as the initial process was repeated. A simple example can be considered to put the problem into perspective. If the target region within the patient has an approximate cylindrical shape with a diameter of 20 cm and a length of 15 cm, 20 elements of the fan-beam of treatment fields (assuming 1 x 1 cm beamlets positioned side by side) would be used to cover the diameter. This fan-beam arrangement would have to be repositioned 14 times to cover the 15 cm length of the target region. If the intensity of the beams is switched every 10 degrees of angular increment as the fan beam rotates around the patient, a total of 720 beams would pass through the patient in one full rotation. Multiplying this number by 15 gives a total number of more than 10,000 separate beams. These are the candidate beams for the optimization process. This is a very large number of beams that must be managed during the optimization and delivery steps.

Binary Multileaf Collimator
In order to implement this idea, a special collimator system, the binary multileaf collimator (MLC), was designed to quickly open and close apertures in front of the different beam elements in the fan beam. Turning again to the example given in the preceding paragraph, the amount of time each element stayed open during the 10-degree angular increment around the patient determines the weight of that beam. The design used in this example is the general design for the tomotherapy IMRT approach used in the device produced by TomoTherapy Inc (Madison, WI) and called the Hi-Art system. This modern version of this technology actually uses a slow, continuous movement of the patient support system with a rapidly rotating x-ray source, so that many rotations are possible in a relatively short period of time using a helical dose delivery technique similar to the approach used in the modern spiral CT scanner. The TomoTherapy device broke away from the treatment unit design that had previously been used in radiation oncology for many years, and the Hi-Art system has the visual appearance of a CT scanner.

Standard MLC
Independent of this work, researchers developed another approach that was built around the radiation therapy treatment delivery tools that were readily available at the time. Nearly all patients in the early 1990s were treated with gantry-mounted linear accelerators that had a long arm that reached out over the patient and could rotate around the body on a bearing positioned beyond the patient's head. One part of this system, however, was relatively new at the time. A number of linear accelerator manufacturers had just released modern versions of the MLC and these devices quickly became a key component of the IMRT dose-delivery process.  includes this component of the system, and shows the shaped treatment field that is automatically obtained with the MLC. The binary device that adjusts the aperture opening time for the fan-beam in TomoTherapy is also classified as an MLC, but the more standard MLC is distinctly different in its design. The standard MLC is a device that consists of tungsten leaves that are thick in the direction of the beam, but narrow in width to give fine definition of irregular field shapes. These leaves are motor driven to extend from one side of the useful beam to the other at speeds between 1 and 3 cm per second. For IMRT dose delivery, the standard MLC could be closed to a single 1 x 1 cm beam element, and this opening could be moved sequentially to different positions in the beam's-eye view projection of the target. Repeating this for a number of different directions for the source of radiation (typically seven to 12 directions are used) gives a very large number of individual beams. For the case of 12 different beam directions and the 20 cm diameter by 15 cm length cylinder used previously, the total number of beams is 3,600. This number is smaller than that in the TomoTherapy example simply because the number of angular orientations is reduced from 36 to 12. The obvious problem with this description is the time it would take to deliver 3,600 individual beamlets in sequence.


Using the Standard MLC
A number of manufacturers of gantry-mounted linear accelerators have included modification that make their equipment ideally suited for using IMRT specifically for the treatment of lesions in the cranium. Novalis (BrainLAB, Westchester, IL) is one example. The approach used by this manufacturer includes many innovations to improve the accuracy of the overall treatment strategy. The important features for the discussion here are the reduced field size opening (10 x 10 cm) and the very narrow (3 mm wide) leaves of the MLC system. The Novalis system has the ability to use this MLC to deliver the dose using the IMRT techniques described in the previous section. Varian (Palo Alto, CA), Siemens (Concord, CA), and Elekta also have gantry-mounted linear accelerators that use this same approach for IMRT dose delivery. These devices, like the BrainLAB equipment, have MLCs with reduced field size that use narrower leaves designed specifically for small-field applications.

Using Circular Invariant Collimators
There are two devices developed for relatively small-field applications that use completely different approaches for directing large numbers of beams through the patient for dose painting. Both of these devices use circular collimators instead of an MLC. One small-field device is the Gamma Knife unit developed by Elekta. The Gamma Knife unit focuses approximately 200 beams through small (depending on the exact model, openings range from about 0.4 to 1.8 cm in diameter) circular collimator openings to a single position within a patient's head. Although some of the beams might not be used because they pass through critical structures as they irradiate the target, the resulting dose distributions are very nearly perfectly spherical. By shifting the focus point to different positions within an irregularly shaped target, the number of beams passing through the target region is increased proportionately. Using, say, 10 different focus points, the total number of beams is approximately 2,000. This number is somewhat smaller, but in the general range of the numbers given for the MLC-based approaches. It is important to point out that the circular collimators used for the Gamma Knife present a different dosimetric problem compared with the square beam elements that fit nicely together for the MLC geometry described earlier in this report. With the Gamma Knife, the basic inverse optimization process remains the same in that a computer program determines the weights for each of the circular beams (working in groups for each focus point) on the basis of dose constraints specified by the radiation oncologist to control dose to both the targets and nearby critical structures. However, as an additional step in the process, the size of the collimators used and the exact position of the small spheres that gives the desired dose distribution with a minimum of missed regions and high-dose overlap regions must be determined. This is referred to as the sphere packing process. Because of the difference introduced by this extra step, this dose painting technique and the one described in the next paragraph are not usually referred to as IMRT. However, they do come under the broader classification of dose painting techniques.

Accuray's CyberKnife is another delivery device that can direct a large number of beams through a patient from different directions. This device has a linear accelerator mounted on a robotic arm. It can move around the patient to direct beams from 100 different node positions located on a partial sphere representing the available clearance of the treatment head. From each of these positions, the treatment head can adopt 12 different angular orientations. Thus, for any optimization, as many as 1,200 candidate beams can be used. This number is considerably smaller than for the candidate beams available for the MLC-based approaches. Like the Gamma Knife, the CyberKnife uses circular collimators (ranging in diameter from 0.5 to 6 cm) that can be changed during the treatment to improve conformality. Also, the standard treatment distance of 80 cm to the point-source of the x-rays can be varied to adjust the collimator size. Many of the candidate beams that intersect the target are not turned on after the optimization is complete so that typical treatments employ less than 150 total beams. However, given this number of beams, the optimizer for the CyberKnife can produce dose distributions that are competitive with those obtained with IMRT using an MLC coupled with a gantry-mounted linear accelerator.

All of the devices discussed in this section were originally designed for irradiation of lesions in the cranium. All have image-guidance features that allow very precise targeting of cranial lesions. However, except for the Gamma Knife, where only the newest model extends the treatment region only a small distance outside the cranium, all are now used for stereotactic targeting of lesions in other parts of the body. Examples are the treatment of lesions in the spine, liver, and lung. These devices are ideally suited for treating relatively small targets in any of these locations. As discussed in the next section, using the two circular collimator devices to treat larger lesions may introduce dose-delivery efficiency problems.


Early researchers recognized the problem of decreased dose-delivery efficiency for IMRT. All of the techniques described herein have reduced dose-delivery efficiency compared with 3DCRT. That is, if invaginations where targets curve around critical structures are ignored, and strictly conformal fields that are adjusted to the outline of the targets, the time the treatment beam remains on will be as short as possible. If the same linear accelerator with the same dose rate is used with an IMRT delivery that irradiates different parts of the target in sequence, the beam will have to remain on for a longer period of time as the dose is painted throughout different parts of the irradiated volume. This problem of decreased dose-delivery efficiency will occur for all of the methods described herein. Dose-delivery efficiency is important because leakage radiation passing through the collimator system and other shielding around the treatment head and reaching the patient increases as the delivery efficiency decreases.

Any of the methods that combine beam elements during dose delivery will at least partially alleviate the dose delivery efficiency problem. Examples are (1) the combining of beam elements into larger fields when using a standard MLC, (2) the irradiation of all of the beam elements in the fan beam together during tomotherapy, and (3) the irradiation of the 200 Gamma Knife beams simultaneously when treating lesion in the cranium. However, a problem remains for the last two examples in that both of these delivery techniques require some shifting of the collimator isocenter relative to the patient to complete coverage of the target. Both of these manufacturers, TomoTherapy and Elekta, have recognized this problem and designed their equipment to minimize leakage radiation. The latest Perfexion model of the Elekta Gamma Knife unit has added shielding aimed specifically at solving this problem. The device produced by TomoTherapy includes extra shielding to partially reduce the leakage radiation problem, and other steps have been taken to make sure that their tomotherapy approach gives an efficiency that is similar to the numbers achievable with a gantry-mounted unit using a standard MLC for IMRT. The CyberKnife unit does not combine beam elements during irradiation, and tends to have the highest leakage of the different methods described herein.

The concern over leakage radiation reaching the patient's total body is related to the possibility that dose painting techniques could lead to an increase in second malignancies. A number of recent publications have addressed this issue.

The art of dose painting using biologic imaging for dose planning is an exciting area of research. Planning, as well as treatment delivery, of dose painting may be characterized by a pointillist approach wherein small dots of radiation dose are painted onto a 3D canvas. The paintbrush can take a number of different forms, and each approach comes with its own set of advantages and disadvantages. A very fine brush will be needed to take full advantage of current and future developments in imaging. However, dose-delivery efficiency must be kept in mind, and linear accelerator design changes might be needed to mitigate the problem of leakage radiation reaching distant parts of the patient's body.