Room: Track 2
To investigate the feasibility of pencil beam scanned (PBS) proton beam tracking for lung tumour treatments using real-time liver-ultrasound (US) guided respiratory motion models.
Multiple-breathing-cycle 4DMRIs of the lung, with simultaneous abdominal 2DUS images, were acquired for five healthy volunteers for 10min each. Deformation vector fields (DVFs) extracted from the 4DMRI (ground truth motion: GTM) were used to generate 4DCT(MR) datasets of two lung cancer patients, resulting in 10 datasets with variable motion patterns. From these datasets, patient-specific motion models, based on auto-regression (AR) and Gaussian process regression (GPR), were constructed by correlating the DVFs to the US images. The AR model was computed for a prediction horizon of 133ms to compensate for system latencies. To test the model, 2-field PBS plans were optimised on the PTV=CTV+2mm of the reference CTs, and 4D dose calculations (4DDC) were used to simulate dose delivery for (a) unmitigated motion, (b) ideal 3D tracking (both beam adaption and 4DDC based on GTM), and (c) realistic 3D tracking (beam adaption based on predicted motion, 4DDC on GTM). Additionally, a ‘confidence gating’ concept, which uses the posterior of the Gaussian process as a ‘gate’ to interrupt tracking delivery when confidence in motion state prediction is low, was tested.
US-guided 3D tracking retrieved acceptable target dose homogeneity, improving CTV D5-D95% (median(standard deviation), averaged over 10 cases) from 45.4(3.6)% (unmitigated motion) to 29.9(2.5)% (ideal) and 30.7(2.9)% (realistic) tracking, compared to 9.5%(1.1) of the static references. Additionally, confidence gated tracking showed a slightly improved dose homogeneity compared to realistic tracking (19.5(1.0)% vs 20.4(1.0)%) for the two analysed datasets.
Our data suggests that US-guided PBS tracking can help restore dose homogeneity for PBS proton therapy in lung tumour treatments without increasing dose to healthy lung. Additionally, ‘confidence gating’ may help improve the tracking accuracy.
Funding Support, Disclosures, and Conflict of Interest: This study was funded by the Swiss Nationaly Science Foundation (SNF).