Thesis (Ph.D.)--Biomedical Engineering, Georgia Institute of Technology, 2007. / Committee Chair: Hu, Xiaoping; Committee Member: Brummer, Marijn; Committee Member: Butera, Robert; Committee Member: Oshinski, John; Committee Member: Sathian, Krish.

Functional magnetic resonance imaging (fMRI) has been widely used to study cortical and subcortical mechanisms related to pain. The pathophysiology of burning mouth disorder (BMD) is not clearly understood. Central neuropathic mechanisms are thought to be main players in BMD. This study aimed to compare the location and extension of brain activation following thermal stimulation of the trigeminal nerve with fMRI blood oxygenation level dependent (BOLD) signal. This study included 8 female patients with BMD and 8 matched pain-free volunteers. Qualitative and quantitative differences in brain activation patterns between the two study groups were demonstrated. There were differences in the activation maps regarding the location of activation, with patients displaying greater BOLD signal changes in the right anterior cingulate cortex (ACC BA 32/24) and bilateral precuneus (pandlt;0.005). The control group showed larger BOLD signal changes in the bilateral thalamus, right middle frontal gyrus, right pre-central gyrus, left lingual gyrus and cerebellum (pandlt;0.005). It was also demonstrated that patients had far less volumetric activation throughout the entire brain compared to the control group. These data are discussed in light of recent findings suggesting brain hypofunction as a key player in chronic neuropathic pain conditions.

Fetal growth restriction (FGR) is a common pregnancy complication resulting in increased neonatal mortality and morbidity. The aetiology of fetal growth restriction is not fully understood, but abnormalities in placental development are, leading to abnormalities in placental structure which are thought to affect supply of oxygen to the fetus. The source of fetal hypoxia is unknown due to the difficulty in obtaining oxygenation data in the context of pregnancy using existing techniques. There is also an absence of data relating to oxygenation in FGR pregnancies. Oxygen-Enhanced MRI (OE-MRI) and Blood Oxygen-Level Dependent (BOLD) MRI permit noninvasive acquisition of data related to changes in the concentration of dissolved oxygen (pO2) and changes in hemoglobin saturation (sO2) under air- and oxygen- breathing (hyperoxic challenge).The aim of this project was to determine whether MRI methods can provide information relating to placental oxygenation in normal and FGR-compromised pregnancy, to investigate fetal brain oxygenation and to assess the potential confound of placental perfusion changes under hyperoxic challenge. After optimization of sequences in non-pregnant volunteers, similar pO2 and sO2 increases under hyperoxic challenge were seen in normal and FGR pregnancy. This suggested placental oxygenation was similar and that fetal extraction of oxygen may be a likelier cause of fetal hypoxia. Normal fetal brain oxygenation was found not to increase under hyperoxic challenge, which may be due to hemodynamic adaptation to limit cerebral hyperoxygenation. Finally, the robustness of these oxygenation results was supported by the lack of placental perfusion changes observed under hyperoxia using Arterial Spin Labeling (ASL).In conclusion, MRI methods successfully provided information on placental and fetal oxygenation in normal and abnormal pregnancy, obtaining novel data informing the aetiology of FGR and the physiology of the fetal brain.

This thesis provides a novel dynamic large-scale network perspective on brain activity of human sleep based on the analysis of unique human neuroimaging data. Specifically, I provide new information based on integrating spatial and temporal aspects of brain activity both in the transitions between and during wakefulness and various stages of non-rapid-eye movement (NREM) sleep. This is achieved through investigations of inter-regional interactions, functional connectivity (FC), between activity timecourses throughout the brain. Overall, the presented findings provide new important whole-brain insights for our current understanding of sleep, and potentially also of sleep disorders and consciousness in general. In Chapter 2 I present a robust global increase in similarity between the structural connectivity (SC) and the FC in slow-wave sleep (SWS) in almost all of the participants of two independent fMRI datasets. This could point to a decreased state repertoire and more rigid brain dynamics during SWS. Chapter 2 further identifies the changes in FC strengths between wakefulness and individual stages of NREM sleep across the whole-brain fMRI network. I report connectivity in posterior parts of the brain as particularly strong during wakefulness, while connections between temporal and frontal cortices are increased in strength during N1 and N2 sleep. SWS is characterised by a global drop in FC. In Chapter 3 I take advantage of rare MEG recordings of NREM sleep to show, for the first time, the feasibility of constructing source-space FC networks of sleep using power envelope correlations. The increased temporal information of MEG signals allows me to identify the specific frequencies underlying the FC differences identified in Chapter 2 with fMRI. The beta band (16 – 30 Hz) thus stands out as important for the strong posterior connectivity of wakefulness, while a range of frequency bands from delta (0.25 – 4 Hz) to sigma (13 – 16 Hz) all appear to contribute to N2-specific FC increases. Consistent with the fMRI results, slow-wave sleep shows the lowest level of FC. Interestingly, however, the MEG signals suggest a fronto-temporal network of high connectivity in the alpha band, possibly reflecting memory processes. In Chapter 4 I expand the within-frequency FC analysis of Chapter 3 to explore potential cross-frequency interactions in the MEG FC networks. It is shown that N2 sleep involves an abundance of frequency cross-talk, while SWS includes very little. A multi-layer network approach shows that the gamma band (30 – 48 Hz) is particularly integrated in wakefulness. Chapter 5 addresses the identified MEG FC findings from the perspective of traditional spectral sleep staging. By correlating temporal changes in spectral power at the sensor level to fluctuations in average FC, a specific type of transient events is found to underlie the strong N2-specific coupling in static FC values. Lastly, in Chapter 6 I make the leap out of the constraints of traditional low-resolution sleep staging, and extract dynamic states of FC from fMRI timecourses in a completely unsupervised fashion. This provides a novel representation of whole-brain states of sleep and the dynamics governing them. I argue that data-driven approaches like this are necessary to fully characterise the spatiotemporal principles underlying wakefulness and sleep in the human brain.