Automatic Insulin Delivery: Artificial Pancreas Controlled by Machine Learning Trained Algorithm Compared to other Therapies for Diabetes Treatment
Author
Other authors
Publication date
2021-04ISSN
2634-1034
Abstract
Abstract
Hypothesis Diabetes Type 1 (DT1) therapy by means of artificial pancreas consisting of insulin pump with continuous glucose monitoring and hybrid "closed-loop" control algorithm trained with machine learning (ML) technology provides better glycemia control than multi-daily injection, insulin pump without continuous glucose monitoring (CGM) and sensor assisted insulin pump therapies.
Methods Using Accu-Chek smart pix software to analyze the data collected in-vivo by JC Peiró, author and DT1 patient, in the period August 2004 to August 2019. AccuChek smart pix has been used to collect 4.621 glycemia tests for a period of 1.241 days. The period measured with continuous glucose monitoring contains data with +90% sensor coverage. Control graphics measure mean and median glycemia, standard deviation, time in range, time above range, time below range, hypoglycemia periods, high blood glucose index and low blood glucose index. In-vivo analysis is validated in-silico using the UVA/Padova T1DMS simulator on a population of 30 individuals of different ages under multi-daily Injection and under "closed-loop" artificial pancreas therapies.
Results Compared to multi daily injection (MDI) therapy, the new artificial pancreas with hybrid "closed-loop" machine learning trained control algorithm reduces 70.7% the periods above range, reduces 67.2% periods below range, time in range% increases 75%, hypoglycemia periods reduce 91.2%, high blood glycemic index reduces 67%, low blood glycemic index reduces 73.8%, median glycemia reduces 20% and glycated hemoglobin reduces 20.1%.
Conclusions/interpretation Therapy for diabetes type 1 using "hybrid closed-loop" artificial pancreas controlled by machine learning trained algorithm provides better glycemic control results than the analyzed therapies of multi daily injection, insulin pump without glucose monitoring and "sensor assisted" insulin pump
Document Type
Article
Document version
Published version
Language
English
Subject (CDU)
004 - Computer science and technology. Computing. Data processing
61 - Medical sciences
Keywords
Pages
6 p.
Is part of
Manchester Journal of Artificial Intelligence & Applied Sciences. Vol. 2 Nº. 1 (2021)
This item appears in the following Collection(s)
Rights
© Manchester Journal of Artificial Intelligence & Applied Sciences. Tots els drets reservats