A Robust-CVaR Optimization Approach to Left-Sided Breast IMRT
H Mahmoudzadeh1*, T Chan2, T Purdie3, (1,2) University of Toronto, Toronto, ON, (3) Princess Margaret Hospital, Toronto, ONTU-G-BRB-4 Tuesday 4:30:00 PM - 6:00:00 PM Room: Ballroom B
Purpose: To test the feasibility of a cardiac sparing IMRT planning approach for patients with left-sided breast cancer using a robust optimization model.
Methods: A robust optimization model was developed for breast IMRT. The concept of conditional-value-at-risk (CVaR) was used in the robust framework to guarantee that the clinical dose volume criteria for targets and organs at risk hold under uncertainty in the patient's breathing pattern. Clinical treatment methods for breast cancer (inhale breath-hold with active breathing control (ABC) or free breathing) were simulated via optimization models. A 4DCT patient dataset with target and organs at risk on each breathing phase was used to simulate a clinical case with a total of 20% increase in lung volume from exhale to inhale over 5 phases. The results of the proposed robust model were compared with those of the current clinical models.
Results: Compared to the conventional IMRT method for breast cancer (with free breathing), the proposed robust-CVaR model resulted in a 14.6% reduction in mean heart dose without compromising the target coverage and dose homogeneity. The clinical dose-volume limits for the heart as well as the clinical target volume were met in robust results. The robust method resulted in 23.9% improvement in the maximum dose to 25cc of the heart volume. The robust results showed very low variability among the quality of planning and realized treatments.
Conclusions: Using CVaR limits in a robust optimization framework can help improve the quality of IMRT treatments. The robust-CVaR can generate a high quality treatment plans, but is delivered during free breathing and does not require patient compliance with an external device. The quality of robust treatment remains the same under irregular breathing. Explicitly including metrics for lung and bigger motion amplitudes in the robust optimization method may further improve the results.