Last Thursday 16/03/2023 at 17:00 pm Mr. Diego Jurado made the reading of the doctoral thesis entitled:"New methodology of analysis of dose distributions by algorithms for calculation of absorbed dose in external radiotherapy that simulate the transport of radiation in medium" directed by Dr. Joan Carles Vilanova Busquets and Dr. Rafael Fuentes Raspall and tutored by Dr. Elisabeth Pinart Navidad.
Different types of dose calculation algorithms for external beam radiation therapy with MV photons coexist nowadays. The oldest ones are convolution/superposition (C/S), which consider all media as water of different densities and report doses in water surrounded by water (Dw,w). Subsequently, advanced algorithms are incorporated, which model in detail the transport of radiation in the medium and allow reporting the dose in medium surrounded by medium (Dm,m) or in water surrounded by medium (Dw,m). Advanced algorithms improve accuracy but, unlike C/S algorithms, their dose values are sensitive to the atomic composition of the medium, which requires new considerations in the analysis of dose distributions and plan optimisation.
To propose a methodology for analysing the absorbed dose distributions obtained with algorithms that model radiation transport in medium that allows using the criteria and parameters established internationally for previous algorithms. To study whether the proposed methodology makes it possible to discern whether the altered dose values are due to tissue composition effects or to a sub-optimal plan.
The doses of advanced algorithms are expressed in such a way that their values are not sensitive to the atomic composition, which formally entails introducing a new dose quantity that we have named dose in reference-like medium (Dref,m*). Dref,m* is linked to a reference medium, and is defined as the absorbed dose in a voxel of that medium surrounded by a medium with the same composition but with the same radiation transport properties as the original. A methodology was developed to derive Dref,m* distributions by post-processing Dw,m or Dm,m and applying a correction factor (CF) to each voxel depending on its composition. The CFs were obtained within the planning system with two virtual dummies of varying density and composition. A proof of concept was performed for the advanced Acuros XB algorithm and a 6-VM beam, comparing the Dref,m*, Dm,m, Dw,m and Dw,w results of the AAA C/S algorithm in several cases. Finally, the performance of Dref,m* and the other quantities in the optimisation process was evaluated for three representative cases that were optimised over the PTV with the criteria established for Dw,w to achieve the same dose distributions. The differences between quantities in terms of radiation fluence, robustness and complexity were analysed.
The values of Dw,w from C/S do not depend on medium composition, but radiation transport may not be accurate enough in some cases. Advanced algorithms solve this issue, but their dose values are sensitive to medium composition: Dm,m presents lower dose values (-1% systematic) and cold spots in bone (-4%) and metal (-10%), while Dw,m presents hot spots (10% in bone and 32% in metal). The developed method to obtain Dref,m* simple and easy to implement. Dref,m* distributions keep radiation transport accuracy and their dose values get rid of these hot/cold zones, thus enabling the evaluation with the established criteria. In optimisation, attempting to compensate for the inhomogeneities of Dm,m and Dw,m introduces fluence compensations that cause patients to be treated differently and inherently worsen robustness. Dref,m* is the most consistent quantity in all cases, obtained with advanced algorithms but without needing extra compensations. This produces naturally robust plans and patients are still treated similarly.
Dm,m and Dw,m values of advanced algorithms are sensitive to medium composition. This sensitivity, different for Dm,m and Dw,m, can alter the dose distributions, impact their analysis and, during optimisation, introduce fluence compensations that affect the doses delivered to the patients and worsen robustness. A new analysis system has been developed to express the dose distributions in terms of Dref,m* so that their values do not depend on medium composition. Obtaining Dref,m* is feasible, enables the use of the analysis criteria established for previous algorithms, and distinguishes whether altered dose values are due to compositional effects or to a sub-optimal plan. Dref,m* allows for robust plans, keeps of doses delivered to patients, enables traceability to clinical experience, facilitates the transition to advanced algorithms, and can be useful in automated planning and multicentre activities. Regardless, the final dose distribution must be reported in Dm,m or Dw,m to correlate with clinical outcomes.