Detection of EEG patterns related to nocturnal hypoglycemia.

The aim of the project was to detect specific EEG patterns related to hypoglycemia. EEG analysis was performed using a probabilistic classifier and unsupervised learning for the construction of learning sets for the classifier. Unsupervised learning and additional tools were used in the search for EEG patterns occurring when the blood-glucose level was below the hypoglycemic threshold. The rate of these specific EEG patterns was below 5% in normal nights. In patients who were known to have no or a reduced glucagon response to hypoglycemia, the rate increased to 20-80%.