Lorenzo Meneghetti1, Eyal Dassau2, Francis J Doyle2, Simone Del Favero1. 1. Department of Information Engineering, University of Padua, Padua, Italy. 2. Harvard John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA.
Abstract
BACKGROUND: Personal insulin pumps have shown to be effective in improving the quality of therapy for people with type 1 diabetes (T1D). However, the safety of this technology is limited by the possible infusion site failures, which are linked with hyperglycemia and ketoacidosis. Thanks to the large availability of collected data provided by modern therapeutic technologies, machine learning algorithms have the potential to provide new way to identify failures early and avert adverse events. METHODS: A clinical dataset (N = 20) is used to evaluate a novel method for detecting real-time infusion site failures using unsupervised anomaly detection algorithms, previously proposed and developed on in-silico data. An adapted feature engineering procedure is introduced to make the method able to operate in the absence of a closed-loop (CL) system and meal announcements. RESULTS: In the optimal configuration, we obtained a performance of 0.75 Sensitivity (15 out of 20 total failures detected) and 0.08 FP/day, outperforming previously proposed literature algorithms. The algorithm was able to anticipate the replacement of the malfunctioning infusion sets by ~2 h on average. CONCLUSIONS: On the considered dataset, the proposed algorithm showed the potential to improve the safety of patients treated with sensor-augmented pump systems.
BACKGROUND: Personal insulin pumps have shown to be effective in improving the quality of therapy for people with type 1 diabetes (T1D). However, the safety of this technology is limited by the possible infusion site failures, which are linked with hyperglycemia and ketoacidosis. Thanks to the large availability of collected data provided by modern therapeutic technologies, machine learning algorithms have the potential to provide new way to identify failures early and avert adverse events. METHODS: A clinical dataset (N = 20) is used to evaluate a novel method for detecting real-time infusion site failures using unsupervised anomaly detection algorithms, previously proposed and developed on in-silico data. An adapted feature engineering procedure is introduced to make the method able to operate in the absence of a closed-loop (CL) system and meal announcements. RESULTS: In the optimal configuration, we obtained a performance of 0.75 Sensitivity (15 out of 20 total failures detected) and 0.08 FP/day, outperforming previously proposed literature algorithms. The algorithm was able to anticipate the replacement of the malfunctioning infusion sets by ~2 h on average. CONCLUSIONS: On the considered dataset, the proposed algorithm showed the potential to improve the safety of patients treated with sensor-augmented pump systems.
Authors: Robert H Slover; John B Welsh; Amy Criego; Stuart A Weinzimer; Steven M Willi; Michael A Wood; William V Tamborlane Journal: Pediatr Diabetes Date: 2011-07-03 Impact factor: 4.866
Authors: Daniel P Howsmon; Nihat Baysal; Bruce A Buckingham; Gregory P Forlenza; Trang T Ly; David M Maahs; Tatiana Marcal; Lindsey Towers; Eric Mauritzen; Sunil Deshpande; Lauren M Huyett; Jordan E Pinsker; Ravi Gondhalekar; Francis J Doyle; Eyal Dassau; Juergen Hahn; B Wayne Bequette Journal: J Diabetes Sci Technol Date: 2018-02-01
Authors: S Shalitin; T Ben-Ari; M Yackobovitch-Gavan; A Tenenbaum; Y Lebenthal; L de Vries; M Phillip Journal: Acta Diabetol Date: 2013-08-24 Impact factor: 4.280
Authors: Andrea Facchinetti; Simone Del Favero; Giovanni Sparacino; Claudio Cobelli Journal: IEEE Trans Biomed Eng Date: 2012-11-15 Impact factor: 4.538
Authors: Dorothee Deiss; Peter Adolfsson; Marije Alkemade-van Zomeren; Geremia B Bolli; Guillaume Charpentier; Claudio Cobelli; Thomas Danne; Angela Girelli; Heiko Mueller; Carol A Verderese; Eric Renard Journal: Diabetes Technol Ther Date: 2016-08-15 Impact factor: 6.118