Compressed sensing-based LDR brachytherapy inverse treatment planning with dosimetric criteria

TitleCompressed sensing-based LDR brachytherapy inverse treatment planning with dosimetric criteria
Publication TypeConference Paper
Year of Publication2015
AuthorsGuthier, CV, Hesser, JW
Conference NameDEGRO
PublisherStrahlenther Onkol (2015) 191:S1–S184
Other NumbersP-10-19-YD

Purpose: Compressed sensing (CS)-inspired solvers for inverse treatment planning (ITP) for low-dose-rate (LDR) brachytherapy yield solutions near the global optimum in real-time. This allows investigating the role of dose volume histogram (DVH) related dosimetric criteria over dose-based objectives for the planning process.

Methods: Treatment planning aims at covering the planning target volume (PTV) with a prescribed dose while sparing organs at risks (OARs). In LDR treatment planning a dose-based objective function models this goal by minimizing the difference of prescribed and achieved dose while penalizing dose differences above and below a given threshold. However, this optimum does not imply optimality with respect to the clinically relevant dosimetric criteria. The direct usage of dosimetric criteria, earlier seen as infeasible with given optimizers, is now in reach with a new CS-based optimizer. Checking clinical usability of dosimetric criteria based objective function, ten patient cases are optimized by three different approaches, a semi-automatic manual forward planning, an inverse planning using the commercial planning system Oncentra Prostate Elekta AB with a state of the art objective function, and our own developed planning system with the CS inspired solver incorporating the criteria based objective function. For the standard dose based objective function a set of optimization parameters that are the standard at the Mannheim University Medical Center are used, whereas for the standard dosimetric criteria for clinical evaluation according to the American Society of Physicist in Medicine are selected.

Results: All three approaches return clinically acceptable plans. For V100 PTV criteria obtained values using the semi-automatic approach are 0.88±0.06 (range: 0.77 to 0.95), 0.93±0.03 (range: 0.84 to 0.95) with Oncentra Prostate, and 0.93±0.02, (range: 0.91 to 0.95) using the new approach. For V150 PTV criteria almost equivalent results for the semi-automatic approach (0.55±0.07, range: 0.43 to 0.64) and the criteria-based approach (0.55±0.02, range: 0.52 to 0.57) are achieved, whereas it was 9.8% lower than Oncentra Prostate solutions (0.61±0.05, range: 0.50 to 0.68). Remaining dosimetric criteria for OARs show no statistical significance. In addition the new approach produces plans, which are more robust with respect to the standard deviation of dosimetric criteria.

Conclusion: A CS-based optimization strategy allows using dosimetric criteria based optimization for LDR treatment planning. Reducing the system complexity from sixteen to six free parameters, it yields an easier system control.

Citation KeyGuthier_DEGRO_2015