depth margins are computed in designing each beam. In the case of scattered beam treatments, the lateral margins would be designed into the aperture in the beam’s eye view, and the depth margin would be designed into the compensator.
For scanned beams, and intensity modulated proton therapy in general, these margins would influence which pencil beams would be used, and the depth of penetration of each one.
When a single beam is used, the beam sizes are enlarged to cover uncertainties in set-up and the beam penetration is increased to compensate for other uncertainties, as mentioned. The issue is more complex when more than one (non-parallel) beam is used. The PTVs for each beam could be reported individually, but they cannot be added because they correspond to different processes.
Regarding the reporting of dose to the PTV, ICRU Report 78 states that:
“it is required that the dose distribution within the ‘PTV’ be recorded and reported. This would be unworkable if there were a separate PTV for each beam employed, and impossible if separate lateral and depth margins were built into the computer’s beam-design algorithm. It is therefore proposed that, in proton therapy, the PTV be defined relative to the CTV on the basis of lateral uncertainties alone. An adjustment must then be made within the beam-design algorithm to take into account the differences, if any, between the margins needed to account for uncertainties along the beam direction (i.e., range uncertainties) and those included in the so-defined PTV (i.e., based on lateral uncertainties)” [20].
The same approach can be extended to carbon ion therapy. For reporting, the anatomical location, extent, volume and dimensions (three orthogonal dimensions) of the PTV should be indicated.
therefore needed to assess the actual dose distribution over a complete course of radiotherapy.
5.4.1. Dose in organs with varying volume
If the target volume or the volume of an OAR varies during treatment, the delivered dose distribution may differ from the planned dose distribution in that volume. These variations may occur rather rapidly, i.e. during a single fraction, or gradually over the complete course of radiotherapy. The effect of volume variations on the dose delivered to that volume depends on the magnitude of the variation, the treatment technique and the margins applied, i.e. the CTV-PTV margin and the margin around an OAR. With the implementation of 3-D CRT and IMRT techniques, and the accompanying trend to reduce these margins, the chances of having areas in the target volume that receive a lower dose than planned, or parts of an OAR that receive a higher dose than desired, are not zero [178].
In order to determine the dosimetric consequences of these variations in volume, the following information is needed: (a) a time dependent description of the volume changes, (b) a calculation of the (static) dose distribution at relevant time intervals and (c) the generation of cumulative dose distributions over the course of therapy of each subvolume of a specific tissue or organ.
5.4.1.1. Time dependent description
As discussed in Section 6.4, a number of volumetric imaging tools are now available to describe the patient anatomy in 3-D either as a function of time during one treatment fraction or during a series of fractions. The main purpose of these tools is to ensure that the position of relevant patient anatomy is the same during each treatment, or just before, compared with the planned position.
By applying in-room imaging, the patient position, the patient treatment or both can be adapted, if necessary, to keep the dose in specific volumes the same within specified uncertainties compared with the planned values. However, if the volume is variable, simple couch position shifts are not always sufficient and more sophisticated methods of using this imaging information are necessary, such as using it to perform image guided adaptive radiotherapy. One of the main problems when dealing with organs with varying volume in fields with steep dose gradients is the tracking of the position of specific subvolumes in that organ.
For that purpose, a number of groups have developed deformable registration algorithms [179, 180]. Furthermore, making a new plan based on the modified position of the PTV and OAR requires a lot of effort in contouring all relevant
structures. In order to reduce that laborious task, automated contouring tools, in combination with atlases of patient anatomy, are under development.
5.4.1.2. (Static) dose distribution
The next step is to calculate the 3-D dose distribution for each time point at which a relevant change in anatomy has been observed. Using the conventional dose calculation procedure by combining a new set of CT data with existing treatment parameters in the clinical treatment planning system would make such a procedure very cumbersome. For that reason, new approaches are under development, such as designing robust treatment techniques that produce dose distributions that are less sensitive to volume variation, or the use of atlases of precalculated dose distributions. A conceptually simple, but in practice still rather cumbersome approach, is the use of ‘plan of the day’ adaptive radiotherapy.
Multiple IMRT plans of a particular patient are generated for various possible positions of the PTV and OAR. By using in-room imaging, the optimal plan of the day is chosen, an on-line set-up correction is applied, and the corresponding treatment plan is irradiated. Other image guided adaptive radiotherapy methods under investigation try to adapt the treatment technique automatically, for instance, by changing leaf positions or collimator angle, to maintain adequate dose coverage of target volume and sparing of OAR. It is obvious that these techniques should be fast to make them clinically useful, while the accuracy should be comparable to results of existing dose calculation algorithms.
5.4.1.3. Cumulative dose distribution
The assessment of the accumulated dose in moving subvolumes of tissue or an organ requires the combination of a time dependent description of the volume variation and a dose calculation for each time point. Various approaches are under development to make such an approach useful for routine application in the clinic [181, 182]. Modelling the movement of the organ due, for instance, to breathing, as a function of time during one fraction, in combination with predetermined dose distributions, might result in a more reliable actual dose distribution than using the planning results. In order to get the cumulative dose distribution, assumptions still have to be made about the constancy of the movement during a series of fractions. The development of deformable registration tools will certainly help in determining the cumulative dose distribution in mobile tissues that only change in position or shape. However, an as yet unsolved problem is how to take changes in the volume and position of subvolumes of these tissues, e.g. of the PTV or of OAR, into account in regions with a large dose gradient.
A detailed discussion of these elements to achieve the determination of the actual dose distribution delivered to normal tissues during a course of radiotherapy has been given by Jaffray et al. [183]. That paper also describes a number of future developments to achieve a high accuracy of the dose within tissues with varying volume.
The impact of dose and volume uncertainties on the generation of dose–
response data has been analysed in recent work by Kurjewicz [184]. Using the Lyman-Kutcher model for lung response, an analysis of 200 virtual experiments was performed, and it was found that uncertainties of 10% in dose and volume resulted in a significant increase in the derivation of the m parameter (slope related parameter) in the model, in addition to yielding large 95% confidence limits in the resulting parameter. The mean value and the 95% confidence levels of the n parameter (volume dependent parameter) were hardly affected.
Better knowledge of the actual dose distribution in the target volume and OAR, incorporating variations in volume in the accumulated dose determination, will allow the assessment of more accurate dose–response curves, and consequently, will yield improved input data for TCP and NTCP models. That knowledge can then be used to design and deliver the optimum treatment to an individual patient.
It should be noted that the total dose map may not be representative for the overall radiobiological effect of the dose, as fractionated dose values can theoretically not be accumulated in a linear way. The effect will be more pronounced for strategies with strong dose variations, for instance, when applying a high dose per fraction. However, in a theoretical study for some typical radiotherapy techniques, Bortfeld and Paganetti [185] showed that a standard deviation of the daily dose fraction of 10% leads to a dose accumulation error due to radiobiological effects of less than 1%. Also, a recent study of the evaluation of the radiobiological impact of anatomical modifications during radiotherapy for head and neck cancer showed that taking radiobiological effects into account while accumulating total dose leads to very small differences compared with a simple linear sum of the dose fractions [186]. For adaptive strategies that make use of the total dose in a limited number of voxels instead of the whole target volume, it might be worthwhile to take radiobiological effects into account.
The level of accuracy of the dose in a volume achievable in practice will be elucidated in Section 6.6, while in Section 7.6.4 the various aspects related to the decision of when to replan a patient will be discussed in more detail.