A Novel Method to Accelerate Optimization by Employing Approximate Dose Values
J Spaans*, H Malhotra, M Ringland, D Nazareth, Roswell Park Cancer Institute, Buffalo, NYTH-A-213AB-1 Thursday 8:00:00 AM - 9:55:00 AM Room: 213AB
Purpose: To introduce a novel method that increases the efficiency of beam and beamlet weight and fluence optimization, by employing approximate dose values in the objective function.
Methods: A least squares objective function was used, based on clinical dose-volume objectives. In the initial stage of optimization, the method utilizes dose values determined by the additive dose approximation (ADA) to evaluate the objective function. These dose values are based on pre-computed dose indices, along with a novel correction scheme. During this stage, no calculation of the true dose is required, resulting in a significant reduction in computation time, since the manipulation of large dose matrices is avoided. The resulting optimized parameters are subsequently used as the starting point in the final optimization step. This method was evaluated on prostate bed, esophageal, and brain patients with varying numbers of treatment fields, by comparing the CPU time required with that of a standard optimization method. The downhill simplex algorithm was used for the evaluations.
Results: Nine prostate bed patients with three different beam configurations were used: 4 field 3D-CRT plans, 16 field 3D-CRT plans and a 64-beamlet IMRT-like plan. These demonstrated an average reduction in optimization time of 33%, 54%, and 46%, respectively. Two 3D-CRT esophageal patients were evaluated, one with 14 fields and one with 16 fields, resulting in an average improvement of 29%. Finally, a 3D-CRT 16 field brain patient was studied and exhibited a 57% decrease in CPU time.
Conclusions: The proposed method, when used in conjunction with existing optimization algorithms, dramatically decreases the number of standard objective function evaluations and, as a result, computation time. This technique may be applied to any type of optimization algorithm and set of clinical variables requiring pre-computed dose matrices and an objective function involving dose-volume objectives (eg. multi-criteria optimization).