Bayesian reconstruction of insulin secretion rate

The rate by which insulin is secreted from the pancreas is important for a quantitative understanding of the glucose regulating system in man. However, the insulin secretion rate (ISR) can unfortunately not be directly measured as the secreted insulin undergoes a large and variable liver extraction before reaching plasma. Nevertheless, a C-peptide is luckily co-secreted with insulin on an equimolar basis which, unlike insulin, is not extracted by the liver and thus the C-peptide concentration in plasma directly reflects the pancreatic ISR. In this project the reconstruction of the ISR is formulated as a Bayesian input estimation problem, which can be solved by deconvolution. The applicability of phase-type distributions in the deconvolution is also addressed. It is our beliefs, that the project may have significant importance for the treatment of diabetics as a better understanding of the ISR may improve the artificial pancreas. Supported by Novo Nordisk A/S