A Bayesian approach to the minimal model

Insulin sensitivity, glucose efficiency and pancreatic responsiveness provide an integrated metabolic portrait of a single individual, which is known to have important predictive properties for the development of diabetes. In this project, a Bayesian method based upon the minimal model for glucose disposal and insulin kinetics of Bergman and his co-workers for the assessment of these three most important factors of glucose tolerance level. Statistical inference about the metabolic portrait is based upon posterior samples obtained by Mar\-kov chain Monte Carlo simulations techniques. By adopting directed graphical models in a Bayesian framework, and hereby imposing an a priori knowledge about the unknown quantities of interest, we have succeeded in efficiently regularizing the ill-posed inverse problem possessed by Bergman's minimal model, when considering the unified model obtained when both glucose disposal and insulin kinetics is modelled. Supported by Novo Nordisk A/S