Program Information
Statistical Process Control to Assess the Quality of Image-Guided Radiotherapy
S Shiraishi*, M Grams , L Fong de los Santos , Mayo Clinic, Rochester, MN
Presentations
WE-F-605-11 (Wednesday, August 2, 2017) 1:45 PM - 3:45 PM Room: 605
Purpose: To develop and validate a statistical process control framework using a similarity metric (SM)-based registration virtual probe for volumetric image-guided radiotherapy.
Methods: 898 CBCT registrations were retrospectively analyzed for 32 bilateral head and neck cancer patients who received definitive radiotherapy. The virtual probe was defined as an 8x8x8cm^3 volume of interest (VOI) around the C2-C5 spine within which the SM was calculated. The sensitivity to systematic misregistrations was analyzed for normalized cross-correlation (CC) and mutual information (MI). This was done by reregistering images to find optimal registrations corresponding to maximum SM values. The variability of SMs was obtained for displacements that varied according to a normal distribution (σ=2mm), and the detectability of displacements greater than 5mm was investigated. Patient-specific control charts were produced using a statistical process control framework. Outliers from control limits were inspected visually. The process was tested on a single patient having a SM mean and variation approximating the patient cohort average to determine if the SM virtual probes could identify two critical failure modes: registration off by an entire vertebral body and registration with incorrect patient images.
Results: Deviations greater than 5 mm were detected at 2.8σ and 2.1σ from the mean for CC and MI probes, respectively. Patient-specific control limits identified suboptimal registrations which were also confirmed by physician review. Group averages for the SM mean and σ were 0.88 and 0.04 for CC and 0.66 and 0.08 for MI. The critical failure modes were identified at 6σ and 4σ by CC and MI probes, respectively.
Conclusion: Cross-correlation identifies suboptimal registrations more effectively than MI due to more gradual sensitivity with respect to displacements. Cross-correlation as a SM is suitable for statistical process control, and offers quantitative information to assess the quality of the volumetric image guidance process in radiation therapy.
Funding Support, Disclosures, and Conflict of Interest: Partially supported by Varian Medical Systems.
Contact Email: