Cone beam CT

CBCT differs from conventional third and fourth generation CT in that the X-rays are not collimated into a fan beam but rather they are collimated into a wide field of view creating a cone of X-rays. A 2D array of detectors, rather than a 1D array as with conventional CT, is required. Amorphous silicon flat-panel detectors are generally used for CBCT as they do not have optical scattering, are well suited to mounting on the linear accelerator and they have high resolution. Projection data acquired during a single rotation of the X-ray tube is reconstructed rather than piecing together incremental slices as with conventional CT. However, there is extra scatter produced due to the wide field of view of the cone beam. MV CBCT uses the equipment available in most radiotherapy departments and only requires additional software. KV CBCT requires new equipment to be attached to the linear accelerator as well.

Verification of position

As patients are planned in 3D, it would be logical to verify their position in 3D also.

Techniques such as intensity-modulated radiation therapy (IMRT) allow a high dose to the tumour but a low dose to normal tissue and therefore there is naturally a steep dose gradient. In fact, a dose gradient of 10% per mm can be achieved quite easily.Reference Chen, Morin, Aubin, Bucci, Chuang and Pouliot¹ With a steep dose gradient accuracy of position becomes even more vital as slight errors in positioning can lead to large errors in the dose received. Two-dimensional portal images can be difficult to interpret because they are not in reference to the 3D CT image sets. Out-of-plane rotations are not as easily seen as translations and although digitally reconstructed radiographs (DRRs) can be used for comparison, when mismatches occur the sources of the error may be unclear.Reference Mackie, Kapatoes and Ruchala²

HanleyReference Hanley, Mageras, Sun and Kutcher³ found that when registering orthogonal portal films to the simulator reference image, errors were introduced to the patient position which had dosimetrically significant consequences for out-of-plane rotations of 2° or more, highlighting the importance of 3D verification. GuckenbergerReference Guckenberger, Meyer, Vordermark, Baier, Wilbert and Flentje⁴ noted large rotational errors even in cases of small translational errors confirming this finding. These dosimetric errors were of clinical significance for patients with elongated non-spherical target volumes and with sharp dose gradients to adjacent organs (e.g., nasopharyngeal cancer with inclusion of cervical and supraclavicular lymph nodes, and the rectum in prostate cancer patients). Therefore, 3D imaging is necessary to eliminate out-of-plane rotations and provide accurate set-up.

One of the difficulties when using CBCT to verify position is that providing manual matching may prove difficult and time consuming and may make an on-line correction policy difficult to implement. Manual matching can also be user dependent. Automated matching would reduce the variability due to user dependence and may also reduce the time required. Automated matching has been found to be capable of detecting known translations and rotations with an accuracy of 1.4 mm for 3D vector offsets.Reference Fox, Huntzinger, Johnstone, Ogunleye and Elder⁵ GuckenbergerReference Guckenberger, Meyer, Vordermark, Baier, Wilbert and Flentje⁴ showed that set-up errors with KV CBCT and automatic registration using bony anatomy differed from electronic portal imaging device (EPID) by less than 1 mm in 70% and less than 2 mm in 90% of cases. Although these errors are not large, a significant number (10%) of cases will have larger errors; therefore, clinical oversight by the radiation therapists would continue to be necessary. This in turn is only possible with adequate training in 3D imaging.

The use of CBCT to verify patient position changes the way in which verification is performed and separates the verification of patient position from the verification of the shape and position of the treatment field. With portal imaging both the patient position and the shape and position of the treatment field are verified together by the one exposure but with CBCT the multi-leaf collimator (MLC) positions and position of the treatment field must be verified separately.

Verification of dose

Most methods of dose verification available at present such as diodes, or metal–oxide–semiconductor field-effect transistors (MOSFETS) which are used instead of diodes in some centres, are point dose measurements. The idea of using the EPID to verify the dose delivered on a daily basis would not only be very practical but also save time on the treatment unit. It would also have implications for verifying more complex dose delivery methods such as IMRT. To develop this concept further, it should be possible to use the EPID system in addition to CBCT to provide a reconstructed 3D distribution of the dose delivered in previous fractions. The treatment plan for future fractions could then be re-optimised to compensate for dosimetric errors. Chen et al.Reference Chen, Morin, Aubin, Bucci, Chuang and Pouliot¹ described a method of reconstructing the dose distribution using EPID and MV CBCT. Before treatment, an MV CBCT scan of the patient is taken and this image is converted to effective photon attenuation coefficients. During the treatment, a portal image of the beam exiting the patient would be acquired. The energy fluence is back projected through the CBCT of the patient, taking into consideration attenuation and the inverse square law in relation to photon fluence. The 3D dose distribution delivered to the patient is then calculated. This theoretically takes into consideration changes in the patient’s shape and changes to the shape of the internal organs along with any dosimetric inaccuracies of the delivered dose, therefore, dosimetric errors could potentially be calculated. There are, however, many difficulties still to be overcome with this method.

First, the energy fluence incident on the detector consists of both primary and scattered radiation which for Chen et al.Reference Chen, Morin, Aubin, Bucci, Chuang and Pouliot¹ led to a 10% higher dose than that predicted from the planning system. Therefore, the energy fluence needs to be corrected for scattered radiation.

Second, Chen et al.Reference Chen, Morin, Aubin, Bucci, Chuang and Pouliot¹ suggests that if structures in the MVCT were contoured, variations from the original plan could be made and under or over dosage of organs could be detected. The accuracy of this delineation however would be very limited as it was noted by SongReference Song, Chiu and Bauman⁶ that the intra-observer variability in contouring of the prostate to be much greater for MV CBCT than for KV CBCT.

Finally, there is a difference exists between the Hounsfield units (HUs) values in the MV CBCT and the original planning KV CT. This is discussed in the next section.

Using MV CBCT, the EPID and beam source positions will move from their ideal isocentric locations due to sagging of the machine. Therefore, there must be a calibration to account for the position of the X-ray source relative to the detector. The same EPID is used for the MV CBCT and the exit beam fluence image, therefore, a positioning mismatch will not exist as the same calibration applies to both. This demonstrates an advantage of MV CBCT over KV CBCT. Therefore if a KV CBCT system is used to acquire the CBCT, a cross calibration of the KV and MV systems would be required to ensure the KV verification system represented exactly the MV delivery system. The idea of using the EPID along with the CBCT to verify the dose delivered should not be discarded but rather requires further development before it is clinically applicable.

Adaptive planning

Adaptive planning is a useful concept where changes in the shape and size of the patient and internal organs over the period of the treatment course can be taken into account. The target organ can shrink, bringing more of the organ at risk into the treatment area. By re-planning during treatment, the treated volume can be changed to match the shrinking tumour. Not only does this provide an increase in sparing of the organs at risk but also a clearer knowledge of the dose delivered. Also, for tumours in such areas as the lungs where fibrosis of the lung tissue during treatment could change the inhomogenity of this tissue, doses could theoretically be re-calculated with new electron densities so the dose delivered would be more accurate.

CBCT has the potential to allow adaptive planning without the necessity of an additional CT appointment over the course of a patient’s treatment. The aim would be to make a modified plan based on the latest information of patient position, target shape and location.

However due to the physical attributes of CBCT, the effects of X-ray scatter and artefacts are larger than in conventional CT.Reference Glover and Pelc^7, Reference Siewerdsen and Jaffray⁸ Using KV CBCT, the use of a bowtie filter attached to the X-ray tube can improve image quality by reducing the variation in the intensity across the detector and charge trapped in the detector.Reference Bushberg^9, Reference Ding, Duggan and Coffey¹⁰ In another study, the use of the bowtie filter with KV CBCT was shown to improve dosimetric agreement with plans based on conventional CT.Reference Yoo and Yin¹¹ This would be expected as improved image quality and reduction of artefacts should improve the dosimetric accuracy.

In phantom studies,Reference Yoo and Yin¹¹ KV CBCT has been shown to have the ability to generate images with HU values comparable with CT. In a patient study,Reference Yoo and Yin¹¹ all tissues except brain in one patient and fat in another showed lower HU values in CBCT than in CT. In a homogenous phantom, the CBCT image was 100–150 HU lower than the conventional CT at the periphery of the image. This discrepancy was between 50 and 200 HU in an inhomogeneous phantom. In actual patients, the HU were also lower which was particularly evident in the back muscles in the vicinity of the spine. The streak artefact contributed to the decreased HU in this region. Noticeable streak artefacts in small volumes also occurred in CBCT images of lung cancer patients because of breathing motion.Reference Yoo and Yin¹¹ The dosimetric consequences of these HU variations was small for the homogenous phantom <1% of MU/cGy. This difference became larger in inhomogeneous phantoms (2–3% in MU/cGy).Reference Yoo and Yin¹¹

KimReference Kim, Horst and Maxim¹² found that in the clinical setting when one breast cancer patient was planned using conventional CT and CBCT differences of approximately 1 Gy were found for the max and mean PTV dose, max patient dose and max lung dose. Bulk density corrections produced results closer to those from the conventional CT plan and KimReference Kim, Horst and Maxim¹² suggests that dose calculations with heterogeneity corrections using KV CBCT HU would not be recommended and bulk density corrections should be done instead. These are large generalisations to make based on one patient although the potential difficulty of the accuracy of HU values in inhomogeneous mediums with KV CBCT may be highlighted for further study.

LoReference Lo, Yang, Schreibmann, Li and Xing¹³ used the same calibration for the conventional CT and the KV CBCT and showed that the difference in HU between the CT and the CBCT was clinically significant. In regions close to the phantom surface, the difference was found to be most significant with differences as large as 400 HU. Approximately, a 3% difference was found between the planning CT and the KV CBCT based planning schemes. The discrepancy between maximum doses rose to 7% in one particular lung patient due mainly to motion artefacts in the CBCT. LoReference Lo, Yang, Schreibmann, Li and Xing¹³ also suggested that mapping the electron density information from the planning CT to CBCT for dose verification calculations might be a solution. This was a small study and the accuracy of electron density values in CBCT again comes into question.

The potential for using KV CBCT to take account of varying homogeneity within the tissues over a course of treatment becomes limited due to the effect of inhomogeneity on the HU values in KV CBCT.

Dose calculation was found to compare well between MV and KV CBCT. Ninety-eight percent of points fell within 3% dose and 3 mm distance to agreement criteria.Reference Chen, Morin, Aubin, Bucci, Chuang and Pouliot¹ In MV CBCT, the quality of the acquired portal image depends on a combination of the ability of the detector to stop the radiation, the strength of the exiting primary beam and the quantity of scatter. MV CBCT of extended objects exhibits cupping artefacts due to the influence of scattered radiation reaching the EPID. If uncorrected, this cupping artefact also appears in the image converted to photon attenuation coefficient leading to errors in the calculated dose. Within a homogenous medium, the dosimetric error was found to be small, approximately 4% for a single open field and therefore a relatively crude correction of the artefact may be acceptable.Reference Chen, Morin, Aubin, Bucci, Chuang and Pouliot¹

Spies et al.’s studyReference Spies, Ebert, Groh, Hesse and Bortfeld¹⁴ found that MV CBCT without scatter correction could lead to estimated electron densities being 30% too low. A scatter correction, using a superposition method with Monte-Carlo (MC) kernels was applied. This improved the contrast of the image by reducing the cupping artefact, and therefore also reduced the discrepancy between the estimated and the actual electron density to 8%. Anti-scatter grids used in KV imaging have only a small effect at MV energies due to the increased energy of the scattered photons.Reference Spies, Ebert, Groh, Hesse and Bortfeld¹⁴ Therefore if the cupping artefact is corrected for in MV CBCT, HU may be relatively accurate in homogeneous phantoms.

MV CBCT which has been shown to register bony anatomy and air cavities with mm precisionReference Xing, Thorndyke and Schreibmann¹⁵ may be useful to register with planning CT but soft tissue visualisation is poor. The effect of inhomogeneous mediums on the HU values has not been investigated to date. However one may hypothesise that the effects of scatter in inhomogeneous mediums with MV CBCT may be less significant than they are in KV CBCT as there is a lower dependence of the scatter on the exact patient anatomy with MV energy. In fact, it has been shown that MV CBCT performs better in the presence of metal objects than conventional KV CT, as high atomic number material has very little impact on the image quality.Reference Morin, Gillis and Chen¹⁶ The impact of metal artefacts, such as fiducial markers, dental fillings or hip prosthesis, in the soft tissue region is magnified in KV CBCT compared to conventional CT because the soft tissue contrast is usually lower with CBCT images. Metal shadows, due to missing information, will introduce streak artefacts which can spread to nearby soft tissue regions in the reconstructed CBCT image. Although this may be corrected for with small metal objects such as fiducial markers, corrections cannot be made for large metal objects such as hip prostheses.Reference Zhang, Zhang, Zhu, Lee, Chambers and Dong¹⁷ In MV CBCT, the image quality is not affected by these metal objects and the dose calculation may be more accurate.

The outstanding benefit of KV CBCT is that the visualisation of soft tissue, and therefore the visualisation of deformity, is better. This allows more accurate contouring for adaptive planning.

Extra dose

The most significant difference between KV and MV CBCT is in the dose required to achieve the same image quality. JaffrayReference Jaffray, Drake, Moreau, Martinez and Wong¹⁸ studied the dosimetric advantage of KV localisation using a phantom where aluminium was used as a substitute for bone. The ratio of MV and KV to achieve equivalent object detection were D6 MV: D90 KVp = 40:1 and D18 MV: D90 KVp = 94:1 indicating that much higher doses were required for MV imaging. The tissue contrast visible at a given resolution and a given dose for a conventional KV CT is superior to that of a MVCT.Reference Ruchala, Olivera, Kapatoes, Schloesser, Reckwerdt and Mackie¹⁹ Furthermore when Groh et al.Reference Groh, Siewerdsen, Drake, Wong and Jaffray²⁰ compared KV and MV CBCT it was found that the increase in contrast and signal-to-noise ratios at KV energies meant that lower doses were required for equivalent detect ability.

The large volume of normal tissue receiving low dose from both KV CBCT and MV CBCT is significant although the dose is much higher with MV CBCT. The risks associated with this low dose to normal tissue include an increased risk of secondary induced cancer. There is some evidence to suggest an increased sensitivity to low levels of radiation and low energy exposures in the KV range.Reference Chen, Morin, Aubin, Bucci, Chuang and Pouliot¹ However, these risks must be weighed against the increased precision and decreased radiotherapy side effects possible with advanced imaging techniques.

Summary

MV CBCT can be performed with standard radiotherapy equipment with additional software, whereas KV CBCT requires additional equipment as well.

• The use of CBCT in 3D positional verification could be particularly useful when detecting out-of-plane rotations. CBCT positional verification also means that MLC position and treatment field position must be verified separately.

• Dose verification can be performed using MV CBCT but there are still several problems that need to be overcome before this method can be put into clinical use.

• With appropriate corrections both KV and MV CBCT can be used for adaptive planning. MV CBCT handles inhomogeneity better however soft tissue visualisation is superior with KV CBCT.

• The dose from MV CBCT is many times larger than the dose from equivalent KV CBCT images.

In conclusion, MV CBCT has several advantages, particularly the ubiquity of EPIDs however patient dose is the major drawback and will remain so for the foreseeable future. MV CBCT may offer a cost effective alternative to KV CBCT once the problems detailed in this article have been solved.