Main Session
Sep 28
PQA 04 - Breast Cancer, Patient Reported Outcomes/QoL/Survivorship, Functional Radiation Medicine, Hematologic Malignancies, Palliative Care, and International/Global Oncology

2746 - Quantification of Error Sources in Stereotactic Arrhythmia Radioablation Treatment Planning and Delivery

03:00pm - 04:00pm ET
Poster Hall - Exhibit Hall A
Screen: 10
POSTER

Presenter(s)

Connor Haberl, MS - University of Ottawa Heart Institute, Ottawa, ON

C. Haberl1, D. Tiberi2, E. Ali3, A. M. Crean4, F. Rajabiyazdi5, A. D. Chan6, C. J. Redpath1, and R. deKemp1; 1Division of Cardiology, The University of Ottawa Heart Institute, Ottawa, ON, Canada, 2University of Ottawa Faculty of Medicine, Ottawa, ON, Canada, 3The Ottawa Hospital, Ottawa, ON, Canada, 4Centre for Cardiac Research Manchester, Manchester, Hong Kong, 5University of Calgary, Calgary, AB, Canada, 6Carleton University, Ottawa, ON, Canada

Purpose/Objective(s):

Stereotactic arrhythmia radioablation (STAR) is an emerging non-invasive treatment for ventricular tachycardia (VT). STAR is guided by multimodal imaging and requires delineation of the arrhythmogenic substrate on the radiation therapy planning CT (RTCT), introducing potential geometric uncertainty. We aimed to quantify geometric uncertainty across four stages of an end-to-end workflow using SHARP-VT, a custom multi-modal STAR treatment planning platform: (S1) image preparation, (S2) target selection, (S3) target transfer, and (S4) treatment delivery motion.

Materials/Methods:

S1 (Image preparation): RTCT, PET CT, and electrocardiographic imaging (ECGi) CT were registered in ten consecutive cases, and the left ventricular myocardium was segmented from each scan. The combined registration and segmentation error was measured using Dice similarity coefficient (DSC) and 95th percentile Hausdorff distance (HD95) of the registered myocardial segmentations.

S2 (Target selection): Five electrophysiologists without prior STAR experience performed target selection in ten patients VT-free at 1yr post-STAR. Inter- and intra-observer variability and accuracy relative to clinical targets were quantified using DSC and HD95.

S3 (Target transfer): Fifteen consecutive targets exported from SHARP-VT as DICOM-RT were imported into treatment planning software and modified to generate ITV/PTV volumes for STAR treatment. Deviation between exported and final targets was quantified (DSC, HD95).

S4 (Delivery motion): Translational (mm) and rotational (deg) corrections made during STAR delivery were retrospectively analyzed in 55 patients immobilized via abdominal compression (made during mid-treatment cone beam CT verification).

Results:

S1: Image preparation yielded a HD95 of 7 ± 2mm (DSC 0.62 ± 0.12).

S2: Intra-user target selection variability yielded a HD95 of 4 ± 2 mm (DSC 0.72 ± 0.13), whereas inter-user variability (HD95 14 ± 9 mm, DSC 0.41 ± 0.16) and accuracy (HD95 17 ± 12 mm, DSC 0.26 ± 0.15) were substantially higher.

S3: Target transfer from SHARP-VT to the radiation treatment planning software and resulting contour refinements resulted in a HD95 of 4 ± 2 mm (DSC 0.82 ± 0.11).

S4: The median [min, max] magnitude of the 3D translational and rotational motion corrections was 0.2mm [0, 0.7mm] and 0.0deg [0, 1.5deg].

Conclusion:

Geometric uncertainty in STAR targeting is introduced across multiple workflow stages, however inter-observer variability in target selection was the predominant contributor, exceeding segmentation, registration, target transfer, and patient motion errors. These findings inform margin justification and highlight the need for standardized STAR planning workflows.