Worst case optimization WCO: a method to account for uncertainties in IMPT

1
Worst case optimization (WCO) a method to
account for uncertainties in IMPT

• D. Pflugfelder, J. J. Wilkens, U. Oelfke

2
Outlook

• Motivation ? previous talk by J. Unkelbach
• Methods
• - Worst case optimization (WCO)
• Results
• - Range uncertainties
• - Setup uncertainties
• - Combination of both uncertainties
• Conclusions

3
Motivation

• In particle therapy, small deviations from the
  planning parameters can result in a considerable
  deterioration of the anticipated treatment plan
  quality
• Uncertainties in treatment planning parameters
  can be reduced, but not avoided. e.g.
• Range uncertainties
• Schaffner et. al. typ. 1 3mm from HU to
  stop. power conversion
• additional CT-artefacts, metal implants.
• Setup uncertainties
• Rutz et. al. PSI positioning system within
  0.5 3mm
• Desireable Treatment plans that are robust
  against (small) uncertainties
• account for uncertainties in the treatment
  planning process.

B. Schaffner, E. Pedroni Phys. Med. Biol. 43
(1998) 1579-92 H. P. Rutz, A. J. Lomax
Strahlenther. Onkol. 181 (2005) 49-53
4
Worst case optimization
Based on the worst case dose distribution DWCO
(see Lomax).
Calculate dose distributions for multiple
scenarios (e.g. minimal, nominal and maximal
range)
Combine them to DWCO
DWCO is a lower bound, not a physically
possible dose distribution.
Similarly, best case dose distribution can be
defined as upper bound ? used for plan
evaluation only
5
Worst case optimization
Optimization
Worst case optimization
Conventional IMPT
Accounting for different uncertainties only
differs in the way DWCO is calculated.
6
DWCO for range uncertainties
Correlate Bragg peaks along each ray
Calculate ray dose distribution for nominal,
minimal and maximal range
Combine to ray worst case dose distribution dWCO
Add dWCO of all rays to total DWCO
7
DWCO for setup uncertainties
Correlate Bragg peaks in each beam
Calculate beam dose distribution for nominal and
shifted positions. (rigid shift by max. setup
error right, left, up, down)
Combine to beam worst case dose distribution dWCO
Add dWCO of all beams to total DWCO
8
Patient Data Para-spinal case
Tumor wrapped around the spinal cord Metal
implants are present Use three coplanar beams (3D
technique, 3mm s, 3mm grid)
9
Results 5mm range uncertainty (rad. depth)
Worst case optimization
Conventional IMPT
Black lines DVHs for 1331 different scenarios
for the range uncertainty
10
Results 5mm range uncertainty (rad. depth)
Beam 1
Beam 2
Beam 3
Upper row conventional IMPT (pw0) Lower row
Worst case optimization (pw1)
11
Results 5mm range uncertainty (rad. depth)
Worst case optimization
Solid black line range increased by 5mm Dashed
black line range decreased by 5mm
12
Results 2mm setup uncertainty
Beam 1
Beam 2
Beam 3
Upper row conventional IMPT (pw0) Lower row
Worst case optimization (pw1)
13
Results 5mm range uncertainty (rad. depth)
setup uncertainty
Setup uncertainty
0mm
2mm
5mm
Beam 2
Beam 2
Beam 2
14
Conclusions

• ? see previous talk by J. Unkelbach
• - IMPT plans can be sensitive to uncertainties
• - Safety margins are insufficient in IMPT
• - Robust plans can be achieved by accounting
  for uncertainties in the optimization
• IMPT plan using WCO was less susceptible to range
  uncertainties than a conventional single-field
  SOBP plan
• Worst case optimization (WCO) and probabilistic
  approach result in qualitatively similar
  treatment plans

Phys. Med. Biol. 53 (2008) 1689-1700