292 - A Ridge-Filter-Based and Group-Sparsity-Regularized (RFGS) Energy Layer Reduction Method for IMPT

Y. Zhu¹, C. Wang¹, B. Lin², A. Wang¹, A. Pin³, M. Alomar⁴, Y. Lin¹, and H. Gao¹; ¹Department of Radiation Oncology, University of Kansas Medical Center, Kansas City, KS, ²Department of Intervention Medicine, The Second Hospital of Shandong University, Jinan, China, ³Ion Beam Applications, Louvain-la-Neuve, Belgium, ⁴Hospital Universitari Son Espases, Palma, Spain

Purpose/Objective(s): Rapid IMPT delivery is crucial for increasing proton therapy efficiency, treating more patients, improving motion management, and enhancing treatment quality by minimizing uncertainties. It also allows for more breath-hold treatments with reduced normal tissue damage by minimizing target margins. This work aims to reduce energy layer switching time, which is a major component of IMPT delivery time, while preserving plan quality, by using a novel Ridge-Filter-based and Group-Sparsity-regularized (RFGS) method for energy layer reduction. Materials/

Methods: The RFGS method introduces three key strategies that are not present in standard IMPT: (1) incorporating ridge filters to broaden the Bragg peak (BP) width, enabling adequate target coverage with a reduced number of energy layers;(2) employing group-sparsity regularization to effectively select combinations of Bragg peaks (both standard and extended), aiming to maintain plan quality while minimizing the number of energy layers; (3) Implementing a multi-criteria optimization algorithm to construct the Pareto surface, which balances plan quality with the number of energy layers. RFGS addresses the treatment planning challenge by considering dose-volume and minimum-monitor-unit constraints, utilizing iterative convex relaxation, block orthogonal matching pursuit, and the alternating direction method of multipliers. These techniques optimize both plan quality and efficiency simultaneously.
Results: The performance of RFGS, which utilizes both standard and extended BP (HYBRID) or only extended BP (RF), was evaluated against the traditional IMPT method that employs only standard BP. HYBRID demonstrated enhanced effectiveness over both IMPT and RF methods, achieving improved target coverage and better OAR sparing. A head-and-neck case is presented in the following table that compares optimization objective value f, max dose of the target, conformal index (CI), mean dose of brainstem (BS), mean dose of body (BD), and number of non-zero energies (NNZ). The dose is in percentage to the prescribed dose.
Conclusion: The proposed RFGS method, which incorporates novel ridge-filter enhancement and group-sparsity regularization techniques, enables a substantial reduction in energy layers while preserving plan quality. This presents a significant advantage over the conventional IMPT approach.

Abstract 292 – Table 1

Model	f	Dmax	CI	BS	BD	NNZ
IMPT RF HYBRID	3.11	117.9	0.66	8.12	4.48	39
	3.17	115.4	0.71	8.17	4.57	38
	2.56	114.4	0.66	6.67	4.37	13/26
IMPT RF HYBRID	4.12	121.2	0.66	8.31	4.57	29
	3.37	120.2	0.62	8.36	4.77	30
	3.15	115.6	0.66	7.40	4.60	11/19
IMPT RF HYBRID	5.99	126.8	0.69	8.71	4.76	20
	4.94	123.9	0.55	8.43	4.89	20
	3.97	119.0	0.63	8.07	4.80	7/13