ESTRO 2024 - Abstract Book

S4194

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

ESTRO 2024

position. Throughout the entire treatment session, the position of a marker block on the patient's abdomen was recorded with an optical camera (RPM, Varian). The RPM signal was used for exhale respiratory gating with approximately 50%-duty-cycle. Post-treatment, the fiducial markers were segmented in all 2D-projections of the CBCTs. Their 3D motion in patient coordinates was estimated using a probability-based method (Fig1) [3] and used as a surrogate for the tumor motion. To estimate tumor motion during treatment, an external-internal correlation model (ECM) was first built to estimate the tumor motion during the CBCTs from the synchronized RPM motion [4]. The ECM was then used to estimate the tumor motion during beam-on from the RPM signal. For patients 1-5, the ECM relied on the post-treatment CBCT. For patients 6-10, a hybrid ECM was constructed, assuming a linear transition between ECMs from the pre- and post treatment CBCTs. The tumor position at the time of each spot delivery was used to produce motion-encoded DICOM plans that emulated tumor motion in beam’s-eye-view as spot shifts and in-depth tumor motion as changes in the proton beam energy [5]. The motion-encoded plans were imported and calculated in the treatment planning system, and the resulting reconstructed CTV doses were compared with the planned dose. The homogeneity index was calculated as HI = × 100. The dosimetric effect of the inter-fractional anatomical changes was investigated by transferring the treatment plan to the control 4DCTs, keeping the original isocenter position relative to the fiducial markers in the exhale phase. The CTV was copied rigidly from the planning CT and the CTV dose was reconstructed both without (static) and with the motion recorded during treatments using the spot-shift dose reconstruction method.

Results:

The mean tumor motion range at treatment was approximately halved with gating compared to the full motion range (Table1). This resulted in root-mean-square errors of 1.2mm (LR), 3.1mm (CC) and 1.6mm (AP) of the tumor position during spot delivery.

Table1. Mean ± standard deviation over all fractions of the tumor motion range, in cranio-caudal direction, and the tumor position error during spot delivery.

Patient

1

2

3

4

5

6

7

8

9

10

Full tumor motion range during beam-on and beam-off periods (mm)

24.4± 2.6

15.6± 2.3

19.8± 1.8

14.8± 2.1

17.1± 2.0

21.6± 2.4

19.6± 2.1

12.9± 2.4

19.1± 3.2

14.7± 1.5

Tumor motion range during beam-on periods only (mm)

11.2± 1.6

6.7± 1.5

9.1± 1.8

6.0± 1.3

6.9± 0.7

11.6± 7.2

8.7± 1.1

7.2± 0.9

9.3± 1.8

7.2± 1.2

Tumor position error during spot delivery (mm)

-1.3± 2.8

-0.7± 2.9

0.8± 3.0

-0.8± 2.6

1.5± 2.1

1.5± 4.7

0.9± 3.3

0.6± 2.5

0.5± 2.6

1.3± 1.9