Purpose: To report on the development of ROSS (Radiation Oncology Smart Scheduling), an automated tool to assist with assignment of radiation therapy plans to planners and daily monitoring for a busy medical physics department with treatments administered over 8 locations and centralized treatment planning (71 planners, 31 linacs, 12 simulators).
Methods: Plan scheduling assigns plans associated to completed simulations at the end of each business day and monitors the assigned plans for unexpected conflicts (start date change, planner absence, etc). ROSS uses constraint programming to generate assignments with no violations (staff credentialing, time-off, clinical conflicts) and optimizes the schedule for equitable distribution of workload among planners. ROSS accesses databases with assigned plans and new simulations to extract relevant information (required credentialing, technique, dates, etc.) and staff schedules. ROSS monitors the databases daily to flag violations (e.g., changes in start dates, new staff conflicts). Constraints and objectives were continuously adjusted and analyzed during the limited clinical rollout since fall 2020 to determine sufficient and optimal choices. Flagged violations were analyzed for a one-month period and divided into high-impact (likely clinical impact) and low-impact (clerical).
Results: Adjustments to ROSS included constraints: limits on the number of plans due daily and in total, ramp-up/down near vacation periods, sufficient availability during planning/fusions, number of new assignments, multi-sites assigned to one planner. Development objectives inserted included: minimizing cross-campus planning, number of fusions assigned, reaching a dynamic target goal in number of plans (defined by clinical duties such as MR linac, same day SRS program, brachytherapy). ROSS’ daily checks flagged 62 (14 high-priority and 48 low-priority) violations. ROSS produces assignments with a clinically feasible schedule that contains 0 violations in scheduling, with average runtimes of 244 seconds.
Conclusion: ROSS demonstrates that automation can meaningfully assist in scheduling and provides early detection of emergent plan assignment issues.
Software, Quality Assurance, Optimization