Future Perspectives

The next step for work with on-board imaging will involve online 3-D/3-D matching. This would enable a genuine 3-D repositioning of the patient as well as offering the best way to make use of the CBCT dataset – CBCT is the only way to gain genuine 3-D information about the anatomy.

Although the clinic is currently limited because the couch only allows four degrees of movement, it is exploring the possibility of gantry or collimator rotation, which could compensate for the problems of the two degrees of motion missing in the basic therapy couch. Another approach considered is the integration of a six-axis couch, allowing a real 3-D repositioning. This solution – in a dynamic repositioning mode – would offer the possibility of realtime compensation for respiration-induced target motion. The integration of such a system is now evaluated in order to overcome the problem of treatment time prolongation due to the duty cycle (beam hold time) being dependent on the individual target volume movement.

Another goal is to achieve online re-planning, for which it is crucial to have automatic segmentation and follow the tumour shrinkage and changes in the structure. With this, an automatic re-contouring of target structure is needed. It is currently possible to conduct re-planning based on the CBCT slices but is it not possible to do it within a process that is three minutes long. At the moment, the online matching is only manual. The data can be exported into Eclipse but that takes time. Offline planning takes 10 minutes, so it would double the treatment time and that is no good – if it takes 10 minutes, this type of re-planning cannot be introduced. At the moment, the CBCT solution is integrated in a way that takes five to eight minutes, which is too long for regular use in routine clinical treatments. Anything that lengthens treatment times unduly must be avoided.

Online re-planning would really mean having the patient on the couch, conducting a CT scan and analysing the scan to judge whether there is a tumour shrinkage. Is the field size still correct or should it be increased or decreased? With these new tools, all the assumptions made so far could become things of the past. If, for instance, lymphoma (which shrinks very quickly) is being treated how does one take that reduction into account for the following sessions? The CT scan takes about one minute and a huge amount of data is collected. As already mentioned, the reconstruction can be accelerated so it takes half the time it takes today, but still more is needed. It is better than nothing, even though the clinic is always one session behind (because it is taking the new situation into account for the following session rather than the current session), but it is still a major step forward from where it was.

Auto segmentation is also a vision for the future – this involves an initial automatic contouring of the target volume depending on whether there is a clear border between the target and other anatomical structure. If not, there is the question of tumour response – it would make sense to integrate this for re-sizing the fields. The clinic would welcome a tool that supports this type of treatment to the target volume by allowing it to be done manually the first time, but afterwards enabling one to analyse the electron density (Hounsfield units) of that target structure and look into the new CT scan to see whether the structure is smaller. This could then be used as the new contour structure for the target volume.