ESTRO 2024 - Abstract Book

S3434

Physics - Dose calculation algorithms

ESTRO 2024

Purpose/Objective:

Ventricular tachycardia (VT) is potentially life threatening and is a result of myocardial fibrosis of the ventricular tissue causing an obstruction in of normal electrical conductivity. A recent and promising approach to treatment for this specific patient population is radioablation using stereotactic body radiation therapy methods (SBRT). SBRT is characterized by ablative, highly conformal doses and aggressive fractionation schemes. In the setting of radioablation for the treatment of VT, patients are treated in a single, 25Gy fraction while the dose in the center of the tachycardia substrate may be allowed to escalate to up to 35Gy. Given the characteristics of SBRT, along with the delicate nature of irradiating a portion of the heart to high dose, accuracy of dose calculation in the treatment planning system (TPS) is imperative. Two common methods of dose calculation that modern TPSs employ are 1) superposition-convolution algorithms like Varian's Analytical Anisotropic Algorithm (AAA) that support fast calculation but can suffer in accuracy in the presence of heterogeneous medium and 2) Linear Boltzmann Transport Equation (LBTE) solver type algorithms such as Varian's Acuros XB that has superior calculation accuracy. While numerous studies have evaluated these two algorithms in various clinical settings, given the novel nature of cardiac radioablation, along with newly emerging multi-institutional clinical trials comparing radioablation against the traditional catheter ablation, it is imperative that an analysis be done to assess the appropriateness of use of each of these types of dose calculation algorithms in this treatment setting. To evaluate each algorithm for dose calculation accuracy in the setting of VT radioablation, forty-eight treated patient plans that were initially calculated using AAA were recalculated using AXB without re-optimization. A third recalculation was conducted using Radformation’s standalone Monte Carlo (MC) algorithm and was taken as the ground truth for accuracy. Both AAA and AXB dose distributions were then compared to MC using 3D gamma analysis with a 1%/1mm criteria and a low dose threshold of 50% of the prescription dose, and a paired T-test was conducted to detect significant differences between gamma agreements. Various dose metrics in the AAA, AXB and MC plans such as PTV and ITV D98% and the plan maximum dose were compared. Organ at risk (OAR) comparisons for each algorithm compared with MC were also conducted including maximum dose to the stomach and esophagus, and mean dose V16Gy to the heart-PTV, and lung V8Gy. Dose differences for AAA and AXB compared to MC were also evaluated using a paired students T-test. Assuming MC as the gold standard for accuracy, we hypothesize that gamma agreement rates will be significantly larger and differences in OAR dosimetry will be lower for AXB vs AAA. Material/Methods:

Results:

Relative to MC, the median agreement rate for AAA was 89.1% and for AXB was 98.3%. The minimum agreement rate over the entire patient cohort was 70.4% and 93.2% for AAA and AXB, respectively. The overall gamma agreement rates were significantly higher for AXB than for AAA (p-value <001). Figure 1 below demonstrates the overall spread in agreement rates with MC for the AAA and AXB cases.