The main risk for patients affected by type 1 diabetes (T1D) is to fall in hypoglycemia, an event which leads to both short and long-terms automatic failure and can be life-threatening especially when occurs at night without subject awareness. Moreover, T1D patients can develop asymptomatic hypoglycemia, reducing the prompt response of the counterregulatory system triggered by the fall in blood glucose. Avoiding hypoglycemia is important in children and adolescents because hypoglycemia episodes may have clinically relevant effects on cognition. Also in adults, cognitive tests assessed that hypoglycemia results in altered cerebral activity, most likely due to the complete dependence of the brain for glucose supply.
The first organ influenced by this fall of glucose in the blood is the brain. Indeed, a lot of studies proved the mirroring of cognitive dysfunction due to hypoglycemia in the spectral power of the electroencephalogram (EEG) signal. In particular, the increase of the power in low frequency EEG bands is a well-known effect during hypoglycemia that seems more pronounced in the EEG recording in the posterior areas of the brain. Pilot studies about the real-time processing of the EEG signal to detect hypoglycemia have indicated that it might be possible to alert the patients by means of EEG analysis. The main advantages in exploiting EEG analysis is that the blood glucose threshold to enter in hypoglycemia has large inter-subjects variations, on the contrary the EEG onset in general occurs before the state of hypoglycemia is critical, i.e., the brain starts to experience neuroglycopoenia and its functions completely fail.
The main aim of this work is to broaden out the quantitative analysis on the altered EEG activity due to hypoglycemia in T1D patients to identify potential margins of improvement in EEG processing and further features sensitive to hypoglycemia. In particular, the analyses are extended to different domains, i.e., time and frequency domains, to deepen the knowledge on the effects of hypoglycemia in the brain. So far, studies in the literature have mainly evaluated these changes only on a single EEG channel level on the frequency domain, but limited information is available on the hypoglycemia influence on brain network dynamics and on connection between different brain areas. To do so, this dissertation is structured in 7 chapters, briefly presented below.
Chapter 1 will start with a brief overview about the impact of T1D and its main effects on daily life. Moreover, the main consequences of hypoglycemia in human brain will be described by reporting the main findings in the literature.
Chapter 2 will present the database where EEG data and blood glucose samples were collected in parallel for about 8 h in 31 T1D hospitalized patients during an hyperinsulinemic – hypoglycemic clamp experiment.
Chapter 3 will address on the main effects of hypoglycemia in the frequency domain. After testing the well-known changes in the spectral power of the EEG signal during hypoglycemia, a multivariate analysis based on the concept of Information Partial Directed Coherence will be presented. In particular, we will confirm the general slowing in the frequency domain and we will show how hypoglycemia affects the EEG functional connectivity.
Chapter 4 will consider the effects of hypoglycemia on EEG complexity. Fractal dimension features, describing both amplitude and frequency properties, will be computed and compared with the results based on Sample Entropy. We will reveal a decrease of EEG signal complexity in the hypoglycemic condition.
Chapter 5 will focus on the consequences of hypoglycemia in the so-called microstates or “athoms of thought”. We will hypothesize that the changes in the frequency domain and the decrease of the EEG signal complexity in hypoglycemia have in common the same resting EEG electric potential amplitude map.
Chapter 6 will describe how hypoglycemia influences the results of cognitive tests, and the relationship between the drop in the tests performance and the EEG quantitative measures presented in the previous chapters. We will find a direct correlation among the changes in the power spectra, the cognitive tests performance and the changes of one resting EEG electric potential amplitude map.
Eventually, Chapter 7 will close the dissertation by interpreting the ensemble of the results from both the medical and engineering point of view, and presenting the possible future developments of this work.