2018/12/01 by Matthew O’Kelly, O'Kelly, Matthew, Aman Sinha +5
Computer Science · Decision Sciences · Mathematics · Medicine · #Advanced Data Storage Technologies #Diabetes Management and Research #Distributed systems and fault tolerance #FOS: Computer and information sciences #Formal Methods in Verification #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Simulation Techniques and Applications #Statistical Methods in Clinical Trials
paper · pdf · doi:10.48550/arxiv.1812.00293
openalex publication_date 2018/12/01 · openalex created_date 2022/08/01 · openalex updated_date 2026/07/28
Modern treatments for Type 1 diabetes (T1D) use devices known as artificial\npancreata (APs), which combine an insulin pump with a continuous glucose\nmonitor (CGM) operating in a closed-loop manner to control blood glucose\nlevels. In practice, poor performance of APs (frequent hyper- or hypoglycemic\nevents) is common enough at a population level that many T1D patients modify\nthe algorithms on existing AP systems with unregulated open-source software.\nAnecdotally, the patients in this group have shown superior outcomes compared\nwith standard of care, yet we do not understand how safe any AP system is since\nadverse outcomes are rare. In this paper, we construct generative models of\nindividual patients' physiological characteristics and eating behaviors. We\nthen couple these models with a T1D simulator approved for pre-clinical trials\nby the FDA. Given the ability to simulate patient outcomes in-silico, we\nutilize techniques from rare-event simulation theory in order to efficiently\nquantify the performance of a device with respect to a particular patient. We\nshow a 72,000\× speedup in simulation speed over real-time and up to 2-10\ntimes increase in the frequency which we are able to sample adverse conditions\nrelative to standard Monte Carlo sampling. In practice our toolchain enables\nestimates of the likelihood of hypoglycemic events with approximately an order\nof magnitude fewer simulations.\n