Patient-Specific Bio-Morphological Features in Cherenkov Imaging for Positioning Verification: A Retrospective Analysis in Accelerated Partial Breast Irradiation (aPBI) VMAT Radiotherapy 📝

Author: Yao Chen, Lesley A Jarvis, Allison Matous, Rongxiao Zhang 👨‍🔬

Affiliation: Dartmouth College, University of Missouri, Dartmouth Cancer Center, Dartmouth Health 🌍

Abstract:

Purpose: Precise patient positioning is critical in accelerated partial breast irradiation (aPBI) to ensure accurate dose delivery to the tumor bed while minimizing exposure to surrounding healthy tissues. This study quantifies positioning variations in a retrospective cohort of aPBI patients using Cherenkov-imaged bio-morphological features and a registration workflow to examine inter-fraction global shifts and locoregional deformations.
Methods: This retrospective analysis included Cherenkov imaging data from 12 aPBI patients, each undergoing an average of five treatment fractions. Patient-specific bio-morphological features, including vasculature, were auto-segmented using a fine-tuned deep learning model. Rigid and non-rigid registration combined workflow was applied to the segmented features to quantify inter-fractional variations including global shifts (absolute positional differences) and locoregional deformations (residual tissue distortions). Observed variations were analyzed relative to typical planning target volume (PTV) margins of 5–10 mm set for aPBI treatments.
Results: The quantified inter-fractional variations across 12 patients included an average global shift of 5.2 ± 3.3 mm and locoregional deformations of 1.9 ± 1.3 mm, with a maximum observed global shift of 10.8 mm. While these variations remain clinically acceptable for some cases, the 5.2 mm average and 10.8mm max global shifts exceed the threshold of standard PTV margins, highlighting the importance of precise patient positioning. This emphasizes the need for advanced surface-guided radiation therapy (SGRT) techniques, including Cherenkov imaging, for accurate online positioning verification before treatment.
Conclusion: This study highlights the presence of setup uncertainties in aPBI treatments and demonstrates the potential of Cherenkov imaging combined with deep learning and registration workflows to quantify positioning variations and enhance positioning precision. The methodology is scalable, offering a framework for future studies on larger datasets to quantify the clinical impact of positioning variations, refine PTV margins, and improve radiotherapy precision.

Back to List