Click here to


Are you sure ?

Yes, do it No, cancel

Performance Study of the Weighted Sum Versus the Augmented Weighted Chebychev Method in Multi Criteria Optimization of Proton Therapy

H Kamal Sayed*, M Herman , C Beltran , Mayo Clinic, Rochester, MN


(Sunday, 7/14/2019) 3:00 PM - 3:30 PM

Room: Exhibit Hall | Forum 7

Purpose: Multicriteria optimization (MCO) of proton therapy has been shown to provide a valuable tool for producing a good quality plans by providing the trade-off information to the planar. We present a performance study of two of the commonly used methods in MCO (the weighted sum (WS) method and the augmented weighted Chebychev method (AWCM)) using two clinical cases

Methods: An adaptive MCO algorithm based on a parallelized evolution algorithm was implemented on a multi-node multi-GPU cluster. The implementation was performed with each of the WS and AWCM methods. Pareto plans were generated for two clinical cases (bilateral head & neck and a pediatric orbit) using the two methods. The clinical dosimetric qualities based on DVH objectives for each method were compared by generating the same number of plans for each of the two methods. Only plans with a good target coverage within 2% (i.e. the D[98%] points are 2% of the prescription value) were used for the comparison. The Organs at Risk (OAR) DVH objectives of the selected plans were averaged and average quantities were used as comparison metric.

Results: The adaptive MCO algorithm was used to generate the same number of plans for each of AWCM and WS methods (150 plans for the head and neck case and 100 plans for the pediatric orbit case). For the pediatric orbit case, the AWCM method showed a better OAR sparing over WS method with up to 12% of lower dose to OAR’s, while for the head and neck case the advantage was up to 32% of lower dose to OAR’s. The superiority of the AWCM method was more prominent in the D[1%] objectives than the D[50%] objectives for both patient cases.

Conclusion: AWCM method showed a dosimetric advantage over WS method for constructing the Pareto surfaces for both patient cases.


Not Applicable / None Entered.


Not Applicable / None Entered.

Contact Email