Memorial Sloan Kettering Cancer Center New York, NY
D. Rose, E. Pinto, J. M. Moran, G. Niyazov, D. Pazgan-Lorenzo, L. Santanam, G. Chow, S. Liu, L. I. Cervino, S. Elguindi, N. Shah, P. Zhang, J. O. Deasy, and A. Li; Department of Medical Physics, Memorial Sloan Kettering Cancer Center, New York, NY
Purpose/Objective(s):We developed and clinically implemented PlanQ, an innovative AI-based approach to managing radiotherapy workflows . The primary objective of this platform is to simplify radiation therapy workflow management, thereby enhancing clinical decision-making. PlanQ achieves this by creating comprehensive digital profiles for patients, which link clinical events across their treatment journey, aiming for heightened efficiency and scalability for diverse user needs in clinical settings.Materials/
Methods: Developed with extensibility in mind, PlanQ supports data integration from multiple systems, enabling the seamless curation of multimodal data in real-time, guided by FAIR (Findable, Accessible, Interoperable, and Reusable) principles. The platform incorporates human feedback for some elements to ensure data integrity. By utilizing advanced natural language processing and a large language model, PlanQ adeptly connects sparse clinical events to form longitudinal patient journeys, facilitating the search and analysis of prior patient treatments due to risks associated with re-irradiation. An AI-based Operational Research model is implemented to manage and optimize treatment planning coordination and scheduling. This model considers various factors such as plan complexity and urgency to efficiently allocate resources among a team. PlanQ was designed to make the clinical workflow visible, enabling real-time management and oversight of the entire treatment planning process. Results: PlanQ has led to operational efficiency improvements across our environment, which consists of 30 linear accelerators. Notably, there was a 70% reduction in the workload associated with treatment planning coordination, a 10% decrease in the workload of planners, and a reduction in plan check time, saving an average of one hour per day. These enhancements are attributed to the platforms efficient data curation capabilities, which streamline the process of accessing patient journeys, thus reducing the time users of different groups spend searching for clinical records and supporting patient safety. The platforms NLP model demonstrated a high accuracy rate of 95% in linking isolated clinical events. Conclusion: PlanQ is adaptable with vendor-neutral architecture and is being developed for compatibility with another electronic medical record system. Through its innovative design and application of AI, PlanQ offered a scalable solution to the challenges of treatment planning and execution, with promising implications for the future of patient care in radiation therapy.
