Room: Room 202
Purpose: 3D optical surface monitoring systems (OSMS) are being investigated as an alternative to RPM for breast DIBH treatment monitoring. Limited by its intrinsic optics and mono-parameter interrogation, OSMS frequently generates irreproducible and even false tracking results. This probability increases significantly when the interrogating ROI is small, flat, and symmetric. Users often interpret these false results as true shifts and act upon them. A fundamental challenge is how to seek true shifts from the noisy and unstable tracking results. Here, we present an evidence-based approach to determine action thresholds for breast DIBH treatment.
Methods: Our evidence-based strategy involves 4 steps: 1) determine the optimal size and shape of the interrogating ROI by using our ROI optimization algorithm; 2) determine the error probability distribution of the tracking results (100 samples per each known shift) for the optimized ROI; 3) calculate key statistics of the error probability distribution, such as the mean and standard deviation; 4) determine the targeted tolerance levels for inspiration deviation, such as Â±3.0 mm in translations and Â±1.0Â° in rotations from baselines; 5) calculate the clinical action thresholds by linearly combining steps 3 & 4 with a 95% confidence interval.
Results: We found that OSMS tracking errors have normal distributions with the means and standard deviations dependent on the size and shape of interrogating ROI. Smaller ROIs yielded normal distributions with larger standard deviations. Larger ROIs produced normal distributions with smaller standard deviations. For the optimized ROI, the clinical action thresholds were determined to be Â±3.7 mm and 1.6Â°, based on targeted tolerance levels of Â±3.0 mm and 1.0Â°.
Conclusion: Due to the close interplay between the tracking error distributions and the size/shape of an interrogating ROI, clinical action thresholds should be computed from the tracking error distributions determined with a clinically meaningful ROI.