[OA029] Optimization of phase space files from clinical linear accelerators. (August 2018)

• Record Type: Journal Article
• Authors: Saxena, Rangoli; Martins, Juliana Cristina; Alhazmi, Abdulaziz; Neppl, Sebastian; Reiner, Michael; Belka, Claus; Veloza, Stella; Parodi, Katia
• Is Part Of: Physica medica. Volume 52(2018)Supplement 1
• Page Start: 12
• Publication Date: 2018-08
• Subjects: Medical physics -- Periodicals; Biophysics -- Periodicals; Biophysics -- Periodicals; Imagerie médicale -- Périodiques; Radiothérapie -- Périodiques; Rayons X -- Sécurité -- Mesures -- Périodiques; Physique -- Périodiques; Médecine -- Périodiques; 610.153
• DOI: 10.1016/j.ejmp.2018.06.101
• Languages: English
• ISSNs: 1120-1797

Abstract

Abstract : Purpose: This project proposes a methodology inspired by the gradient descent algorithm[1] to manipulate existing IAEA phase space files (PSF) of an Elekta Precise linear accelerator (linac) head, in order to produce a new fine-tuned PSF for an Elekta Synergy linac head. Methods: An Elekta Synergy coupled with an Elekta Agility multi leaf collimator was modeled in Geant4. Due to unavailability of detailed geometric information of the Synergy head, the IAEA PSF for the Elekta Precise[2] linac was used as initial benchmark. As the geometry of the aforementioned linac heads differs, discrepancies between measured and simulated PDD and lateral profiles in water were quantified in terms of cost value. The Elekta Precise PSF was divided into a grid of squared elements in the x-y plane, each of them holding information on respective particles' initial parameters, such as energy and momentum directions. For each element, small perturbations to the parameters were introduced independently and interim PSFs were generated. Profiles were simulated using the perturbed PSFs and the cost values were updated, yielding a correlation between the cost function and the perturbation. The parameters that produce the minimum cost value were then found iteratively. Results: The influence of particles' energies and momentum on the cost function has been evaluated. By finding a minimum value of cost, optimal parameters can be determined. The final PSF containing optimal values is being validated and an overview of the results will be presented. Conclusions: A methodology for optimization of existing PSF was developed, enabling the generation of fine-tuned, machine-specific PSF without requiring knowledge of linac's head geometry.