ESTRO 2024 - Abstract Book

S4396

Physics - Intra-fraction motion management and real-time adaptive radiotherapy

ESTRO 2024

First, the MLC leaf and jaw tracking performance was evaluated by tracking a sinusoidal trajectory with a squared field in leaf direction (CC) or in collimator direction (left-right (LR)/anterior-posterior (AP)). CC speeds were maximally 12cm/s and LR speeds 8cm/s. The performance was qualified by the root-mean-squared-error (RMSE) of the actual leaf position with respect to the requested position. During the dosimetry experiments, the QUASAR MRI 4D motion phantom (IBA QUASAR, London, ON) was used with a movable insert containing a Ø3 cm spherical gross-target-volume (GTV). The phantom was positioned on a ramp inclined at a 20° angle about the LR axis to decompose the phantom motion in CC and AP motion. To also simulate LR motion, the ramp was angulated by 10° around the anteroposterior axis (Fig 1). The phantom was first used without motion (static), and then programmed with cardiorespiratory motion (cardiac: cos 4 , 70BPM, A peak-to peak =10mm; respiratory: cos 4 , 12BPM, A peak-to-peak =20mm). The phantom angulation decomposed the motion into CC (A peak-to-peak =28mm), AP (A peak-to-peak =10mm), and LR (A peak-to-peak =5mm) components. The GTV motion was monitored using sagittal 2D cine-MR (13.3Hz) with an angulated imaging plane (AP=10°, RL=20°), effectively aligning the imaging plane with the translational motion direction of the phantom, such that the 3D GTV motion could be captured using a 2D imaging plane. A single 2D imaging plane was necessary to achieve the high cine update rate, which was imperative for tracking rapid cardiorespiratory motion. To compensate for the system latency in the MLC tracking workflow, a linear ridge regression predictor was used with variable look-ahead (225-300ms) depending on current delay. Prediction was performed on the combined cardiorespiratory motion, and the accuracy was quantified by the RMSE between the predicted and the actual position. The predicted positions were decomposed into 3D coordinates to track the actual 3D phantom motion with the MLC. Two treatment plans were created for the phantom: a 2000MU squared field (5x5cm2, gantry angle=0°), and a 15 beam 1x25Gy IMRT cardiac radioablation plan with 3mm GTV-to-PTV margins following our clinical template. The latter plan was MU-scaled to 6.25Gy to limit the delivery time. To dosimetrically validate the MLC tracking performance, a cassette insert with 8 PSDs (Medscint, Quebec City, QC) was placed in the motion phantom. The PSDs provide accurate time-resolved measurements at 15Hz with instant dose read-out [3]. One PSD was positioned in the center of the GTV. The other PSDs were positioned around the edges of the GTV on a single line following the translational axis of the phantom.