ESTRO 2024 - Abstract Book

S4279

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

ESTRO 2024

Respiratory gating may reduce the impact of motion during proton therapy of liver cancer, but consistency between external gating marker motion and internal target motion from planning CT to treatment is crucial. Unfortunately, tools for on-line motion monitoring are limited and intra-treatment fluoroscopy is typically not available. This study demonstrates the use of projections from daily setup CBCTs for investigating (1) setup accuracy, (2) motion variation from planning to treatment and (3) consistency of external-internal gating levels.

Material/Methods:

13 patients received 58 GyRBE or 67.5 GyRBE gated pencil beam proton therapy in 15 fractions. For each patient, this study includes data from the 10-phase planning 4DCT and the setup CBCTs acquired at 3-5 fractions (56 CBCTs in total). The target (iCTV) was defined on the exhale phase of the 4DCT as the union of the CTVs in the five phases closest to max exhale (50% duty cycle). SFUD robust optimization with ±5 mm (LR/AP), ±7 mm (CC) shifts and ±4.5% range uncertainty was applied. Daily setup was performed by manually matching the exhale position of 2-3 implanted target-near markers in a free-breathing CBCT to the marker positions in the planning CT (Fig. 1, top-right). Gating was guided by an external abdominal marker block (Varian RPM) with an amplitude gating window corresponding to 50% duty cycle around exhale (Fig. 1). Retrospectively, the fiducial marker motion during CBCT was extracted by marker segmentation in the raw 2D CBCT projections (~700 projections per CBCT), followed by 3D motion trajectory estimation by a probability-based method (Fig. 1) [1]. A small tungsten sphere attached to the RPM block and visible in the CBCT projections provided synchronization between CBCT and RPM data. (1) Setup accuracy. The exhale position (e.g. position 0 mm in Fig. 1) of the fiducial marker constellation centroid was defined from the motion trajectories as the mean 3D marker position during the time the markers were inside 90th 100th percentile position in the CC direction. Comparison with the planned marker positions provided the optimal couch shift for exhale alignment. The difference to the online match provided the online match error. (2) Motion variation. Each individual respiratory cycle was identified in the marker trajectories (Fig. 1). The motion range during the full CBCT and within individual respiratory cycles was compared with the single respiratory cycle captured in the planning 4DCT. Full motion amplitudes and 50th percentile levels (corresponding to 50% duty cycle exhale gating) were compared. (3) External-internal gating level consistency. The synchronized RPM and CBCT data provided the centroid internal marker positions when the external RPM block was inside a 50% duty cycle gating window (Gate-ON). We calculated the fraction of simulated Gate-ON time spent more than 0 mm, 1 mm and 2 mm outside an optimal internal gating window defined as the 50th percentile position in the planning 4DCT (corresponding to the motion included in the iCTV, Fig. 1).

For clarity, only motion along the cranio-caudal (CC) direction was reported in this abstract.

Results:

(1) The mean (± SD) online marker constellation centroid match error was 0.0 ± 0.4 mm (LR), 0.8 ± 0.9 mm (CC), -0.4 ± 0.6 mm (AP).

(2) Large variation between motion during 4DCT and treatment (CBCT) was observed. Limiting motion to the 50th percentile position markedly reduced both motion magnitude and variation relative to the planning 4DCT (Fig. 2, Table 1).