S Decker¹*, D Alexander¹, P Bruza^{1, 2}, R Zhang^1,3,4, E Chen⁵, L Jarvis^3,4, D Gladstone¹, B Pogue^1,2,6, (1) Thayer School of Engineering, Dartmouth College, Hanover, NH, (2) DoseOptics, LLC, Lebanon, NH, (3) Norris Cotton Cancer Center, Dartmouth-Hitchcock, Lebanon, NH, (4) Geisel School of Medicine, Dartmouth College, Hanover, NH, (5) Cheshire Medical Center, Keene, NH, (6) Department of Medical Physics, University of Wisconsin-Madison, Madison, WI

• 
  S Decker

MO-G-202-6 (Monday, 7/11/2022) 2:45 PM - 3:45 PM [Eastern Time (GMT-4)]

Room 202

Purpose: During radiotherapy, Cherenkov imaging provides the unique opportunity to detect unwanted incidents by visualizing the beam on a patient in real-time. This study focuses on the automatic detection of both inter- and intra-fractional patient positioning errors by using various image similarity metrics to compare the positions of inherent biological markers, such as blood vessels, in the recorded Cherenkov images.

Methods: Phantom studies were first completed with an anthropomorphic chest phantom with artificial vasculature simulated for a breast radiotherapy treatment. It was systematically shifted in each translational direction to mimic patient motion, and Cherenkov images were captured during treatment for each position. Mutual information (MI) and the gamma passing rate (%GP) were compared to existing field shape matching image metrics, the Dice similarity coefficient and mean distance to conformity (MDC), to assess how each reacted to patient motion. Additionally, patient cases of known incidents, including positional setup anomalies and intrafractional patient non-compliance, were analyzed to determine the best choice in comparison metric for a given incident.

Results: MI and %GP detected phantom positional shifts as the vasculature moved relative to the beam edge, degrading monotonically as the shifts increased in magnitude. For patient cases, all metrics readily detected intra-fractional motion during beam on. Additionally, MI and %GP can detect intensity differences that may be caused by bolus-tissue air gaps. However, for inter-fractional setup anomalies causing gross beam shape misalignment, the currently employed metrics (Dice and MDC) remain the best choice for detection.

Conclusion: Using the appropriate similarity metric, various inter- and intra-fractional radiotherapy treatment incidents can be detected via Cherenkov imaging analysis. Once anomaly thresholds for each metric are determined, these results will be used to develop an automatic incident flagging algorithm, allowing for a more efficient method of detection compared to the current, time-consuming manual review process.

Funding Support, Disclosures, and Conflict of Interest: This work has been funded by NIH grant R01EB023909. Brian Pogue, Petr Bruza, and Lesley Jarvis are affiliated with DoseOptics LLC, which provided hardware support for this study. Savannah Decker and Daniel Alexander are research consultants for DoseOptics LLC outside the context of this work.

Keywords

Not Applicable / None Entered.

Taxonomy

IM/TH- Image Analysis (Single Modality or Multi-Modality): Computer-aided decision support systems (detection, diagnosis, risk prediction, staging, treatment response assessment/monitoring, prognosis prediction)

Contact Email