Are you an EPFL student looking for a semester project?
Work with us on data science and visualisation projects, and deploy your project as an app on top of Graph Search.
For type 1 Diabetes, insulin injection therapy is the most used treatment today, and insulin pumps are surely the most accurate therapeutic strategy. The insulin use temperature has to not exceed 37°C to prevent unwanted degradation and loss of function. The JewelPUMP (TM) developed by Debiotech S.A., is directly patchable on the skin patient and was the first of its type to integrate a temperature sensor. However, it is unacceptable to place this sensor in direct contact to the insulin, regarding sterility integrity and the actual design of the device. The present study aimed to define how the insulin temperature could be predicted based on the values recorded by this sensor. Two setups were designed to assess the problem in vivo and in vitro: the first includes a patient wearing the pump under different environmental conditions, and the second is based on an electrical thermoregulation system mimicking the human body and its environment by placing it in a climatic chamber. The in vivo setup allowed us to propose a specific Curve Fitting Model (CFM) matching to the real data, based on heat transfer laws and assessing the dynamic of the problem. Two main heat sources were identified: the patient body and the ambient air around the pump. We determined that the insulin restrained in the reservoir reacts more to the body temperature Trbody than to the ambient temperature Trout; inversely, the sensor is more impacted by Trout than by the Trbody. At steady state, the presence of an offset between the sensor and the real insulin temperatures was confirmed. This offset is positive when Trout > Trbody, and negative when Trout < Trbody. Under ideal conditions, if Trout and Trbody are constant, a constant offset was supposed. The behaviour of this offset was investigated with the in vitro setup. The in vitro setup allowed us to monitor accurately the different parameters of the problem. We validated that our thermoregulation system, based on a Peltier system, was able to mimic the thermal behaviour of the human body and the human skin in vitro. Regarding the dynamic, as for in vivo experiments, the proposed CFM matched well to the obtained data. At steady state, the offset was always positive for Tlout > Tlbody, negative for Tlbody > Tlout, and equal to 0 when Tlout=Tlskin corresponding to the equilibrium of our thermal system. The offset after stabilization (Offsetlstab.) was determined as to be a constant exhibiting a dependence on the difference between Tlout and Tlbody. For a fixed Tlbody, the Offsetlstab. increases with Tlout; for a fixed Tlout, the Offsetlstab. increases with Tlbody. This offset ranges between -5°C and +5°C. Based on in vivo and in vitro data analysis, we were able to affirm that it is impossible to predict the real insulin temperature, based only on the single temperature sensor. We identified two main solutions in order to solve this problem. The first consists to placed a different sensor closer to the insulin solution. The second is based on the idea to add a second sensor somewhere else in the JewelPUMP (TM).
Dolaana Khovalyg, Mohamad Rida
Sandrine Gerber, Luca Szabó, Léo Bühler