749 / 881
Conclusions
•
QA of deformable image registration is complex
•
Deformable image registrations is unsolved problem;
algorithms lack biological and biomechanical knowledge
•
Sliding tissue
•
Tumor growth and regression
•
This is OK to make pretty pictures and propagate OAR
contours
•
This is
not OK
for dose accumulation: it is unsafe to estimate
you know where previous dose went
•
This is
not OK
for adaptation around ‘shrinking’ tumors
•
I therefore strongly suggest no to optimize dose on top of
‘accumulated’ dose