Unencrypted login | home

Program Information

Beam Selection for 4Pi Non-Coplanar Converging (4PiNC) Delivery Using a Genetic Algorithm (GA)

D Cho

D Cho*, S Wang, A Sabbas, K Chao, J Chang, Weill Cornell Medical College, New York, NY

SU-E-T-645 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

To develop a GA for selecting 4PiNC beams that optimize the planning target volume (PTV) coverage and organ at risk (OAR) sparing.

A simple treatment planning system was implemented for studying 4PiNC beam delivery. In this system, 4PiNC beams were delivered in a conical fashion where the cone angle was the angle between the central axis of the beam and the transverse iso-plane. We used an ideal spherical phantom with a radius of 10 cm. The 4PiNC beams consisted of 263 beams on 9 (0°, ±20°, ±40°, ±60°, ±80°) cones. The PTV was a sphere (radius=1.4cm) centered at the isocenter and the OAR was another sphere (radius=2cm) 3mm apart from the PTV. The dose in each voxel was calculated using the output factors, tissue-maxima ratios and off-axis ratios of a 6MV beam. An GA was developed to select the beams by minimizing a cost function of important factors including conformity index and maximal OAR doses. For each generation, the GA produces a number of offspring by randomly turning on or off a group of randomly selected beams from the parent. The cost function for each offspring is evaluated and the offspring with the minimum cost function is selected as the parent for the next generation. This process is repeated until all preset constraints are met.

For the PTV only case the GA selected 165 out of total 263 beams that minimized the conformity index to 1.00. With the OAR, 130 out of total 263 beams were selected for producing the best conformity index, 1.10 while maintaining the maximal OAR dose within acceptable levels.

We have developed a GA for 4PiNC beam selection and demonstrated that a subset of the 4PiNC beams are sufficient for optimal PTV coverage and OAR sparing.

Funding Support, Disclosures, and Conflict of Interest: This work was partially supported by a DOD grant DOD W81XWH1010862.

Contact Email: