Room: Exhibit Hall
Purpose: To formalize and automate quality assurance (QA) in radiation oncology. QA in radiation oncology entails a multi-step verification of complex, high-technology treatments involving an inter-disciplinary team and multi-vendor software and hardware. We addressed the pre-treatment physics chart review (TPCR) using formal methods derived from computer science to dynamically track the solution state and automatically identify logical inconsistencies and how they propagate.
Methods: We used a modular approach to decompose the TPCR process into tractable units comprising sub-processes, modules and variables. Module-associated variables served as inputs to the sub-processes. Modules included "Treatment Planning" and "Record and Verify". Sub-processes included "Dose Prescription", "Documents", "CT Integrity", "Anatomical Contours", "Beam Configuration", "Dose Calculation", "3D Dose Distribution Quality" and "Treatment Approval". Dependencies and conflicts between variables were captured in a dependency matrix. The checks performed and sequence in which sub-processes and variables were visited was described in an activity workflow diagram. The solution state was tracked by assigning states to the variables and sub-processes. We converted the TPCR into a Satisfiability Modulo Theory (SMT) problem and used solvers to detect inconsistencies automatically.
Results: The model for the TPCR process comprised 5 modules, 17 sub-processes and approximately 300 variables (~70 distinct). As proof of concept, we formulated the dependency matrix, activity workflow and checks on a reduced model comprising 11 processes and 70 variables (~30 distinct). We identified 107 dependencies and 1 conflict from the dependency matrix. After using an SMT solver, we identified complete paths through all the variables with a minimal number of dependencies, conflicts and stop-start cycles. We were able to automatically detect variable inconsistencies early and check how changes to one or more variables affected different stages of the process.
Conclusion: Constraint solvers from computer science hold promise for formalizing complex QA processes in radiation oncology and automating the TPCR process.