Heijmen Ben Fully automated treatment plan generation in daily

Department of Radiation Oncology
Fully automated treatment plan generation
in daily routine
Heijmen B, Voet P, Dirkx M, Sharfo A, Rossi L, Fransen D, Penninkhof J,
Hoogeman M, Petit S, Mens J-W, Méndez Romero A, Al-Mamgani A,
Incrocci L, Breedveld S.
8th European Conference on Medical Physics, Athens, 2014
Current treatment planning: an iterative trial-and-error procedure in which
the dosimetrist tries to steer the TPS towards an acceptable plan by
tweaking of parameters, such as objectives or weights
 Plan quality is strongly dependent on skills and experience of the
dosimetrist (operator dependence).
 Plan quality is dependent on allotted time.
 Difficult to decide when to stop; could more time result in a better plan?
 Plan generation often based on templates (non-individualized)
 Plan generation may take from 30 min. up to many hours
 Due to involved workload, planning is costly
An alternative: Automated Treatment Planning with
Erasmus-iCycle: Med Phys. 2012; 39(2): 951-963.
Monaco: Elekta AB, Stockholm, Sweden
Prostate wishlist
For each treatment site the wishlist is a priori established
iterative procedure of
Erasmus-iCycle plan
followed by wishlist updates
Institutions can generate their own wishlists
main features of Erasmus-iCycle/Monaco:
 under the hood: lexicographic, multi-criterial optimization in
Erasmus-iCycle, using priorities of objective functions
as defined in the wishlist
 Erasmus-iCycle optimizes fluences, Monaco takes care of
 planning is fully automatic (‘push button system’, no tweaking) 
huge reduction in planning workload and result is operator
 works for IMRT and VMAT, IMRT plans are Pareto optimal
 for IMRT Erasmus-iCycle has automated selection of beam angles
(coplanar and non-coplanar beam arrangements)
How good is automated planning compared
to ‘manual’ planning?
Head and Neck cancer
Int J Radiat Oncol Biol Phys. 2013; 85(3): 866-72.
Study design
• On average 1 in 5 patients got an automatic plan next to the regular
clinical plan made by dosimetrists
• Dosimetrists and treating physicians didn’t know whether or not there
would be an automatic plan.
• All plans were coplanar and had a maximum of 9 beams.
• Treating radiation oncologist selected the plan for treatment
Study results
• in 32/33 cases automatic plan was selected by physician
(almost always better sparing, often also better tumor coverage)
• also objectively (DVHs, NTCP) automatic plans had higher quality.
differences between automatic and manual planning in
mean OAR doses
How good is automated planning compared
to ‘manual’ planning?
Prostate cancer
How good is automated planning compared
to ‘manual’ planning?
Prostate cancer
Study design
30 previously treated prostate cancer patients (78 Gy, IMRT/IGRT)
VMAT plan, automatically generated with Erasmus-iCycle/Monaco
1. IMRT plan, manually generated with Monaco in clinical routine,
and actually delivered
2. VMAT plan, manually generated with Monaco by expert planner
in absence of time pressure
VMAT automatic
IMRT manual
VMAT automatic
VMAT manual
expert planner
no time pressure
Conclusions for prostate cancer
With automated planning:
- Higher plan quality than in clinical routine
- No loss in plan quality compared to an expert planner in
absence of time constraints (non-clinical condition)
- ALWAYS: vast reduction in workload
Same conclusions for 44 cervical cancer patients
 Compared to ‘manual’ planning, automated planning with
 has higher or non-inferior plan quality
 plan quality is not operator dependent
 plan quality is not dependent on allotted time in a busy clinic
 has negligible workload
 Automated planning is currently in clinical use for prostate cancer,
head-and-neck cancer, and cervical cancer.
Next step
International validation study of automated planning with
Erasmus-iCycle/Monaco, together with Florence, Leeds, Mannheim,
Vienna, and Elekta AB.

similar documents