Documente Academic
Documente Profesional
Documente Cultură
Autumn 2017
1. Introduction
radiotherapy department is one such solution that has allowed for repetitive and routine
tasks to be automated, therefore speeding up the planning process and reducing user
errors. Using forward planning techniques many of the tasks can be automated, Mitchell
et al, 2017 investigated automation techniques into breast radiotherapy planning
practice through commercially available auto-segmentation software and non- complex,
non-commercial scripting solutions within a commercial treatment planning system.
Scripts have also been created to be used by the planner to check for common errors
in an effort to reduce the incidence of re-plans. And, finally, they have been used to
optimise the plan itself allowing for auto-planning. Wang et al, 2016 concluded in their
paper that clinical inverse treatment planning process can be automated effectively with
the guidance of an assemble of prior treatment plans and that this approach has the
potential to significantly improve the radiation therapy work-flow.
2. Methods
well-defined process which begins by loading a class solution which automates PTV
growth, plan structure generation, isocenter placement and the definition of optimisation
parameters. After loading the class solution the algorithm fully automates the
optimisation of the plan by dynamically modifying planning objectives within Pinnacle.
Planning objectives are modified by passing out objective dose, volume, weighting and
composite objective values to external Perl scripts, which then analyse the data using
a simple decision tree that effectively mimics the human planner decision process. The
algorithm leverages the availability of Pinnacles composite objective values. Each user-
specified planning objective has its own objective value, which is a measure of how
close the dose distribution is to meeting that objective. The algorithm compares the
magnitude of each objective value, relative to the total objective value to determine if an
objective requires modification. Objectives with relatively small values will have a more
stringent set of dose, volume and weighting parameters applied; and objectives with
relatively large objective values will have a less stringent set of parameters applied. The
results of the algorithms analysis are used to update the planning objective list within
Pinnacle. Each time the planning objectives are modified the plan is reset and re-
optimised with the new parameters. After a few optimisations the algorithm converges
on an optimal set of planning parameters, resulting in a plan which automatically has
favourable organ at risk dose statistics.
3. Results
The development of the ARTS algorithm led to the review of our departments prostate
protocol and the tightening of organ at risk planning objectives. The dose objectives
given in Table 1 represent our current clinical protocol and contains average dose-
volume statistics for 30 clinically acceptable ARTS VMAT prostate plans. The statistics
Automated Real Time Scripts in Clinical Treatment Planning 4
figures/QAflow.PNG
As part of the development process 10 prostate patients have been dual planned
by Human planners and the ARTS algorithm. Table 2 includes average dose-volume
statistics for all 10 patients. In general the ARTS algorithm shows slightly better organ
at risk dose statistics whilst sacrificing a small amount of PTV coverage. In general the
difference between the Human and ARTS dose statistics are small, meaning that the
ARTS generated plans are at least equivalent to their Human planned counterparts.
Automated Real Time Scripts in Clinical Treatment Planning 5
figures/splash.PNG
4. Discussion
The ARTS algorithm has been commissioned for use in our department and has been
Incorporated into the VMAT prostate planning work-flow. All prostate plans are
initially planned using the ARTS algorithm and later assessed by an experienced VMAT
planner. Any sub-optimal plans are adjusted and re-optimised before being released for
checking.
Table 2. Average organ at risk dose-volume statistics for 10 prostates dual planned
by both a Human planner and the ARTS algorithm.
Organ Dose Objective Volume Objective(%) ARTS Average(Gy) Human Average(Gy)
PTV PG V64.2Gy Zero 62.10 61.46
PTV PG V57Gy >99 99.75 99.81
PTV SV V49.4Gy >99 99.90 99.83
Rectum V24.6Gy 80 63.13 61.63
Rectum V32.4Gy 70 46.10 46.7
Rectum V40.8Gy 60 30.19 31.46
Rectum V48.6Gy 50 18.44 19.46
Rectum V52.8Gy 30 10.50 11.16
Rectum V57Gy 15 5.30 5.71
Rectum V60Gy 3 0.70 0.90
Bladder V40.8Gy 50 20.93 22.05
Bladder V48.6Gy 25 15.04 16.04
Bladder V60Gy 5 2.74 1.92
or more complicated plans, can cause an issue for the algorithm because of its limited
ability to compromise planning objectives. In such cases the algorithm generates sub-
optimal plans which require further modification by experienced planning technicians.
Aside from providing a significant time saving, the use of ARTS also frees up a
planner to complete other tasks whilst the algorithm is running, therefore increasing
output and efficiency. ARTs is currently being developed for other VMAT treatment
sites including rectum, bladder and head and neck. Protocols for DMPO plans are also
being extended along with ARTS into a new planning work-flow utilising scripting. This
will allow the plan to be created and finalised in a couple of mouse clicks.
One of the challenges experienced was the lack of a naming convention used in the
department which made scripting difficult. As the department moved towards more
script based planning, naming conventions began to be used as this allowed planners to
use the scripts.
Automated Real Time Scripts in Clinical Treatment Planning 7
5. Conclusion
Theoretically, each patient planned by the algorithm will receive the same quality
of plan, wherein each planning objective is achieved and each individual organ at risk
receives its minimal possible dose. Currently, the algorithm can achieve planning goals
for 90 percent of standard prostate plans; and further developments, based on user
feedback, will increase the algorithm’s overall efficiency.
The relative success of the original algorithm has led the development of a
generalised version of ARTS, which is capable of planning multiple sites with multiple
prescriptions, independent of the planning objectives specified by the planner.
6. Acknowledgements
References
Round, C. E. et al. (2013) Radiotherapy Demand and Activity in England 2006-2020, Clinical Oncology,
25(9), pp. 522-530. doi: 10.1016/j.clon.2013.05.005.
Mitchell, R. A. et al. (2017) Improving the efficiency of breast radiotherapy treatment planning using a
semi-automated approach, (March 2016), pp. 18-24. doi: 10.1002/acm2.12006.
Wang, H. et al. (2017) Development of an autonomous treatment planning strategy for radiation
therapy with effective use of population-based prior data, Medical Physics, 44(2), pp. 389-396. doi:
10.1002/mp.12058.