Abstract
We describe how scale space methods can be used for quantitative analysis of blood glucose concentrations from type 2 diabetes patients. Blood glucose values were recorded voluntarily by the patients over one full year as part of a self-management process, where the time and frequency of the recordings are decided by the patients. This makes a unique dataset in its extent, though with a large variation in reliability of the recordings. Scale space and frequency space techniques are suited to reveal important features of unevenly sampled data, and useful for identifying medically relevant features for use both by patients as part of their self-management process, and provide useful information for physicians.