Developing and clinically testing a learning algorithm that optimizes the performance of the artificial pancreas

在健康个体中，血糖水平由两种激素严格控制：胰岛素和胰高血糖素。1型糖尿病患者胰岛素分泌减少，胰岛素是降低血糖的重要激素.因此，目前1型糖尿病患者通过注射或胰岛素泵进行强化胰岛素治疗，但血糖控制仍然存在问题，对大多数患者来说是一个很大的挑战。我们小组正在稳步研究开发一种体外人工胰腺，可以自动调节1型糖尿病患者的血糖水平。人工胰腺由三部分组成：连续葡萄糖传感器（测量血糖水平）、输液泵（输注胰岛素和胰高血糖素），and an intelligent智能dosing剂量algorithm算法（根据传感器读数控制胰岛素和胰高血糖素输送）我们已经在我们的临床研究机构进行了几次临床试验，结果很有希望。人工胰腺显著改善了葡萄糖，与传统胰岛素泵治疗相比，我们目前正在国内进行更长时间和更大规模的实际研究，该项目的目的是改进人工胰腺，使其能够随着时间的推移了解患者的胰岛素需求，并适应患者胰岛素需求的变化。我们相信，这种新的算法将保持和提高系统的性能，面对患者胰岛素需求在数周/数月内的缓慢变化。我们的目标是在计算机模拟环境中测试这个新系统，然后在50名青少年中进行为期3周的临床试验。人工智能具有巨大的潜力，可以改善生活质量，改善健康状况，预防低血糖事件，为体力活动提供信心，减轻糖尿病患者，家庭和医疗保健提供者的负担。

In healthy individuals, blood sugar levels are tightly controlled by two hormones: insulin and glucagon. Secretion of insulin, avital hormone to reduce blood sugar, is lost in type 1 diabetes. Thus, type 1 diabetes patients are currently treated byintensive insulin therapy through injections or insulin pump, but glucose control remains problematic and a big challenge formost patients.Research is under steady progress by our group to develop an external artificial pancreas that automatically regulatesglucose levels in patients with type 1 diabetes. The artificial pancreas is composed of three components: a continuousglucose sensor (to measure glucose levels), infusion pumps (to infuse insulin and glucagon), and an intelligent dosingalgorithm (to control insulin and glucagon delivery based on the readings of the sensor).We have tested the artificial pancreas in several clinical trials in our clinical research facility and the results were promising.The artificial pancreas improved significantly glucose control and reduced the risk of hypoglycemia compared toconventional pump therapy. We are currently moving to longer and larger real-life studies at home.The aim of this project is to improve the artificial pancreas so that it learns the insulin needs of the patients over time andadapts as the patients insulin needs changes. We believe that this new algorithm will maintain and improve the performanceof the system in face of slow changes in patients insulin needs over weeks/months. We aim to test this new system in acomputer simulation environment then clinically in 50 adolescents for 3 weeks.The artificial has enormous potential to improve quality of life, improve health outcomes, prevent hypoglycemic events,provide confidence for physical activities and alleviate key aspect of the burden of diabetes for patients, families and healthcare providers.