UC Irvine

ABSTRACT OF THE DISSERTATIONIdentifying Quantitative Biomarkers in the Electroencephalogram for the Evaluation of Developmental and Epileptic Encephalopathies By Derek Kuan-Yu Hu Doctor of Philosophy in Biomedical Engineering University of California, Irvine, 2023 Associate Professor Beth A. Lopour, Irvine, Chair

Developmental epileptic encephalopathies, such as infantile epileptic spasms syndrome (IESS) and Lennox-Gastaut Syndrome (LGS), are severe epilepsy syndromes that begin early in life. These syndromes are characterized by drug-resistant seizures, which can cause developmental delays and lifelong impairments. A favorable clinical outcome is reliant on a prompt diagnosis, but this is complicated by the variability in individual cases and the evolution of the epileptic encephalopathy into other refractory epilepsies. The electroencephalogram (EEG) is an indispensable tool to help evaluate these syndromes, as it provides a noninvasive measure of time-varying neural activity and can contain waveforms that are indicators of epilepsy. Furthermore, the application of computational techniques, such as functional connectivity and time-frequency analysis, can help identify and quantify the features of these waveforms, providing an objective measurement to guide clinical diagnosis and decision making. Therefore, we aimed to develop quantitative EEG biomarkers for IESS and LGS for three purposes: (1) to improve diagnosis by providing clinicians a normalized baseline for healthy development, (2) to evaluate the functional network changes associated with pathological EEG waveforms, and (3) to discover EEG biomarkers that cannot be reliably detected using visual review. First, we measured the changes in EEG functional connectivity networks during normal development in the first two years of life. We showed that an increase in network strength was typical during normal infant development, while the network structure remained stable. This work can serve as a baseline for future investigations in IESS, which is associated with altered functional connectivity networks. Second, we assessed the functional connectivity changes in patients with IESS during interictal epileptiform discharges (IEDs), a common epilepsy-associated waveform seen in EEG recordings. We found an increase in connectivity strength during IEDs that could be attributed to the underlying pathophysiology of the disease, rather than an increase in connectivity due to the broadly distributed waveform. Lastly, we developed a novel method for the discovery of EEG biomarkers using unsupervised analysis of the time-frequency decomposition. We applied this method to patients with LGS and confirmed that it could detect established EEG biomarkers of the disease. It also highlighted novel candidate biomarkers that occurred more frequently in patients with LGS than healthy controls but were not recognized as salient waveforms during visual review by epileptologists. A separate study confirmed the low interrater reliability for classifying EEG waveforms associated with LGS, which emphasizes the need for this type of objective, automated algorithm. In all, this work advanced computational EEG techniques, such as functional connectivity and time-frequency analysis, to complement standard visual review. In the long-term, incorporating such methods into clinical practice can lead to improvements in diagnosis and patient care which were previously not possible.