Session Item

Complexity of intensity-modulated radiotherapy (IMRT) plans depends on e.g. the applied optimization procedure, and a high degree of multi-leaf collimator (MLC) modulation is known to challenge the accuracy in both dose calculation and treatment delivery. In automated treatment planning for online adaptive radiotherapy (oART) plan quality evaluation relies on calculation-based approaches, and the use of MLC complexity metrics could act as a valuable supplement to independent dose calculations. This study aimed at evaluating two different MLC complexity metrics in comparison with the more commonly used MU/Gy for clinically generated oART plans on a linear accelerator with a double-stacked MLC.

A total of 57 clinical IMRT plans for seven patients treated for urinary bladder cancer were exported from Ethos (Varian Medical Systems); a CBCT-based and artificial intelligence-driven oART platform with a double-stacked MLC. The MLC complexity metrics, edge penalty [1] and one minus the modulation complexity score (1-MCS) [2], were calculated from the control points in the DICOM files, and information on the MU/Gy was collected for each plan. Phantom-based measurements were carried out using the Delta4+ phantom (Scandidos) on the corresponding plans and the resulting gamma passing rates (3%/2mm, local, 20% cut-off) were compared to the complexity scores. The correlations between the gamma passing rates and the various metrics were evaluated.

The Pearson’s linear correlation coefficients between the gamma passing rates and the edge penalty, 1-MCS, and MU/Gy were r=-0.83, r=-0.80, and r=-0.74, respectively. All metrics demonstrated a negative correlation with the gamma passing rate, with slightly stronger correlations for the MLC complexity metrics than with MU/Gy. Furthermore, the presence of an outlier in the data for MU/Gy (Figure 1C, patient 6) indicates a potential risk of false negatives when evaluating plan quality using only this metric. Such outliers were not observed with the MLC complexity metrics (Figure 1A and 1B), indicating a potential clinical value of defining cut-off thresholds for plan quality using these metrics.

The gamma passing rates from the phantom-based measurements has superior correlation with the MLC complexity metrics edge penalty and 1-MCS compared to the more commonly used method of evaluating the MU/Gy, as a measure for plan quality. The 1-MCS and edge penalty complexity metrics could therefore potentially serve as a valuable supplement to calculation-based QA of online adaptive IMRT plans, where a phantom-based measurement is not an option.

[1] Younge KC et al. Penalization of aperture complexity in inversely planned volumetric modulated arc therapy. Med Phys. 2012;39(11):7160-70.

[2] McNiven AL et al. A new metric for assessing IMRT modulation complexity and plan deliverability. Med Phys. 2010;37(2):505-15.