Linear accelerators (linacs) can now deliver high-energy beams, with conventional linacs employing C-arm gantries. These beams can be geometrically shaped, using a multileaf collimator to conform geometrically to the beam's-eye view of the target. Add intensity modulation of the radiation beam (IMRT) to the mix, and high-dose distributions with concave outlines can be created.

Other innovative techniques include slice-based and helical tomotherapy, the Accuray CyberKnife robotic radiosurgery system (see image on homepage) and new gimballed linac heads. Some researchers are even contemplating IMRT using cobalt radiation sources.

What's more, advances in computer and detector technology mean that image-guided radiation therapy (IGRT) is now a reality. Three-dimensional imaging apparatus for kilovoltage and megavoltage cone-beam tomography, together with the exploratory use of ultrasound, linac-linked MRI and a variety of optical techniques, form the basis of such IGRT.

So with all of these options available, why haven't we cured cancer? That's a big question. While other specialists work to target cancer with drugs and gene therapy, radiation therapy remains a key way to cure many tumours, to arrest the growth of others, and to palliate and improve quality of life. And it's a technique that's likely to be needed for another half century, maybe more.

Underlying issues

Radiation-therapy physicists sensibly attack one problem at a time. In recent decades the goal was to create concave, high-dose distributions using IMRT, to show experimentally that they can be delivered, and to roll out the technology to as wide a hospital base as possible in order to start clinical treatments and trials. The maximum dose that can be delivered is limited by both the target tolerance and any failure to conform to the target.

Today we can safely say that we are very good at treating phantoms, tissue-equivalent materials and patients who do not move (much). Providing we side-step the resource issues that lead to lower-than-desirable clinical availability, this has been a triumphant success.

But herein lies the rub - patients do move. Even if a patient's skin contour appears stationary, the organs within can, and do, move. This motion arises from respiration, cardiac motion, peristalsis and digestive processes that alter the contents of our waste-disposal organs.

Traditionally the issue of patient motion has been solved by treating what is known as the planning target volume (PTV). This includes not only the region where the clinical target volume (CTV) is during imaging and planning but also where it might be during the treatment. In the absence of any further knowledge, the margins invented for that are no bad thing.

Indeed, elegant mathematical work has recently shown that the concept of margins even works for IMRT, the "jigsaw on a jelly" technology, where at first sight it would seem that it shouldn't. The Central Limit Theorem comes to the rescue, and all is fine because radiation therapy is fractionated.

However, in the ideal case we don't actually want to treat the PTV, because in doing so we over-irradiate normal structures. Instead we want to treat the smaller, mobile CTV - but how? Readers must not expect a single answer. Research in this area is relatively new, but fortunately, when physicists spot an interesting problem, they flock to it like vultures to carrion. Thus conferences are currently awash with talks on this topic and the literature is burgeoning.

Taking control

The most important factor when dealing with patient movement is establishing the size of the problem. X-ray imaging of implanted metal markers can do that, as can GPS-like imaging of implanted radio emitters. Sadly, we have learned that tumours move differently depending on the tumour site. And even tumours of the same kind move differently in different patients. Worse still, the motion of a given tumour in a given patient can vary from moment to moment.

So is it simply time to sigh and give up? Thankfully, no. We can look instead for ways to control the motion. Some patients can be coached to breathe regularly, perhaps by watching a marker that tracks their breathing against a computer-generated trace. Others can be taught to breath-hold throughout treatment fractions or have their breath-hold controlled by active spirometer technology. When attempting to solve the motion problem, I would say that the first task is to minimize it by such means. In fact, without regularizing the breathing, I believe that treating the CTV by any technique would remain beyond us.

If one could genuinely know the target motion, then gating the therapy to a small fraction of the radiation-on duty cycle would minimize, if not solve, the motion problem. The plus side of such a gating technique is its relative simplicity. The down side is that it increases treatment time. And ultimately, gating is only ever as good as the accuracy with which the target motion is known.

This brings us nicely to another major concern. We cannot sensibly use X-rays to view tumour motion throughout therapy fractions because of the unwanted increase to the radiation dose that this would cause. Instead we employ surrogates, which are markers that are usually placed on the skin.

Surrogate monitoring might include detecting the motion of an infrared marker (active or passive), the capture of fringe patterns or reflected light from large areas of the skin surface, and radio-markers. Less precise surrogates include monitoring surface motion through a pressure sensor or monitoring the airflow into and out of the lungs (either directly or via air temperature). It's important that future research addresses the issue of understanding (and preferably controlling) the relationship between motion of the internal target and that of the external marker.

The alternative to gating, and a solution that enables continuous radiation delivery, is tracking. The goal here is simple: to arrange the collimator motion such that the target is stationary in the frame of the moving radiation. This is, however, easier said than done.

Volumetric targets do not move as rigid bodies, so the problem of tracking differential motion arises (to which there is no perfect solution). In addition, elastic motion changes tissue density, which affects dosimetry, and there is usually a latency between detecting motion and adjusting collimation. Work is ongoing on all of these problems, but it's too early to implement this type of treatment.

So what's the bottom line? Four-dimensional CT and MRI are beginning to show patterns of organ motion, but the technology is nowhere near close enough to routine implementation for a single solution for organ motion to be proposed. Thus, parallel research tracks must be followed to develop gating, breath-hold and tracking methods. It may be the case that none of these will be ideal for all clinical situations and a menu of techniques will be needed.

For anyone going for treatment in the near future, breath-hold and gating could be at least discussed with the clinic, as they are patchily available in the UK. Tracking isn't, but it's the solution that occupies most of my current research time, and one that I hope will one day be available. Others in my department are active on many of the other possibilities.