Session Item

This study proposes a model based on new complexity indexes compilation to predict the feasibility of treatment plans in radiotherapy with intensity-modulated.

The clinical model was created in three stages. First, development of an algorithm in python 3.7, integrating the calculation of 7 complexity scores: MCS / LTMCS [1] and MAD / MFA / CLS / CAS [2]. The scores describe parameters such as moving and opening leafs for each control point (CP). An ad hoc score has been developed to integrate the impact of Monitor Units Variability (MUV). Then, a retrospective study evaluated the correlation between the scores and the Gamma Indexes (GI) of the patient-QA from a database including 1379 VMAT beams. To validate the model, a prospective study was carried out on 451 VMAT beams. The model has been established to identify plans that are likely to fail the GI test as recommended by AAPM [3], [4]. An home made software has been developed.

A strong correlation (R = 0.79) was found with the LTMCS / MUV combination and the GI (3% 3mm) for the normo-fractionated planes and (2% 2mm) for the stereotaxic planes. The highest found value of LTMCS is 0.57 (low complexity) and the lowest value is 0.02 (high complexity). The MUV varies between 0 (UM stable between all CPs) and 180 (Maximum variation of UM between CPs). The results showed that a LTMCS greater than 0.15 and a MUV less than 5 of a plan mean that its beams are predicted as true positives (GI> 95%). Sorting by the LTMCS and the MUV made it possible to have 80% (1104 beams out of 1379) of true positives (p-value <0.001). The scores calculated for the 20% of the remaining plans (275 beams) reflect a high complexity between each CP, and may lead to GI <95% (70 out of 275 beams). The method combining the LTMCS / MUV allowed to discriminate the risk of false positives (p-value <0.001). The prospective use of the model proved its detection efficiency of 80% of true positives (359 out of 451 beams, p-value <0.001), of the remaining 92 beams, only 25 were true negatives (GI <95%).

The new decision-tree integrating MUVscores provide help with patient-QA controls. The two LTMCS / MUV scores help rule out the false positives. Specifics models including the complexity scores and the thresholds are dependent on each TPS / Linac pair. Work in progress to reduce the 15% of false negatives.

References :
[1] Jong Min Park (2015). Modulation index for VMAT considering both mechanical and dose calculation uncertainties. Physics in Medicine & Biology.60
[2] Crowe (2014). Treatment plan complexity metrics for predicting IMRT pre-treatment quality assurance results. Australasian Physical & Engineering Sciences in Medicine, 37(3)
[3] AAPM Task Group No. 119, IMRT Commissioning Tests, Instructions for Planning, Measurement, and Analysis
[4] AAPM Task Group No. 218, Tolerance limits and methodologies for IMRT measurement-based verification QA