ESTRO 2024 - Abstract Book

S3584

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2024

5 Centro Diagnostico Italiano, Cyberknife Department, Milano, Italy. 6 Mater Dei Hospital, Città di Bari Hospital spa, Cyberknife Department, Bari, Italy. 7 A.O.U. Policlinico G. Martino, Medical Physics Unit, Messina, Italy. 8 Fondazione IRCCS Istituto Neurologico Carlo Besta, Health Department, Milano, Italy

Purpose/Objective:

Conformity and steep dose fall-off are fundamental requisites for intracranial stereotactic treatments. This study analyzed inter-institute dose gradient variability of CyberKnife (CK) brain SRS/SRT plans. The aim was to investigate the feasibility of simple predictive models to better guide and possibly automate intracranial SRS/SRT planning optimization

Material/Methods:

Seven CK centers joined the present study developed within a broader Italian Multi-Institutional project. Three hundred clinical plans, created for brain single lesions distant from serial critical structures and treated between 2018 and 2023, were retrospectively enrolled and analyzed. A minimum of 30 to a maximum of 64 cases were enrolled per participant. For each case, geometrical/anatomical, dosimetric and planning details were reported. PTV median volume was 1.44 cc (range 0.06 to 29.9 cc) and lesions were treated in single (211), three (73), four (3), and five (13) fractions schedules with dose prescriptions ranging from 14 Gy to 40 Gy. Centers were equipped with different CK systems: one S7, two M6, two VSI, two G4. Plans were created for Fixed (148), Iris (136) and MLC (16) collimator using different optimization algorithms (VOLO (228), Sequential (39) and Simplex (33)). Prescription dose, minimal prescription isodose, planning optimization approach were specific for each single center. For each plan, conformity index (CI) and Dose gradient index (DGI) were estimated according to Reynolds 1 formula. According to this definition, the higher the DGI value the steeper the dose fall-off: DGI values ≥ 100 corresponds to a gradient ≤ 0.3 cm, whereas any additional millimeter corresponds to a loss of 10 DGI points. The equivalent radii of volumes defined by a set of isodose lines (100%, 85%, 50%, 30%) were also calculated 2 . Data analysis focused on the correlation between PTV (volume and radius) and parameters quantifying the dose fall off. Patients were stratified into six group by PTV size: Group I (0-1 cc), II (1-3 cc), III (3-5 cc), IV (5-10 cc), V (10-15 cc), VI (15-40 cc). For each group, the DGI average value was defined as “Ideal” and the minimum acceptable DGI (Minimal) as the value rejecting the lowest 10% of cases. To determine the Minimal values a normal distribution of DGIs was assumed. The analysis was repeated restricting to plans with a conformity index (CI) < 1.2 considered an ideal condition for brain SRS plans 1 . CI and DGI median values were respectively 1.09 (0.96-2.23) and 84.77 (range 19.62-108). DGI as a function of PTV volume is shown in figure 1. A strong significant correlation (p<0.0001) was found between PTV size and DGI with Pearson’s r= -0.78 (95% Confidence Interval: -0.82 to -0.73). DGI ideal values were 95, 82, 77, 71, 59 and 52 for group I, II, III, IV, V and VI respectively. The corresponding minimal DGI values were 86, 73, 67, 59, 43, 36. The rate of DGI passing the minimal criteria was about 90% independently of the group. Considering each center separately, the passing rate with respect to the multicenter Minimal values, ranged from 43% to 100% over the six groups. The inter-institute variability of DGI, expressed in terms of standard deviation, increased with target volume from 6.7 for group I to 12.47 for group VI. Restricting to plans with CI<1.2 (n=220), ideal DGI values remain within 1.4% of the values calculated over all cases. Radii of volumes included in 100%, 85%, 50% and 30% isodose lines are shown in figure 2 as a function of PTV radius together with the R 2 values for the corresponding linear regression models. The linear fitting showed a good accuracy. R 2 ranged from 0.958 (30% isodose) to 0.994 (100% isodose). When considering only plans with CI<1.20, R 2 values increased uniformly (0.966-0.997). Results: