Author:

Abstract:

Epilepsies are one of the most common neurological disorders, and seizures are their primary clinical manifestation. From the clinical point of view, several challenges remain in patient diagnosis, treatment, and follow-up. From the scientific standpoint, epilepsies represent a conundrum that requires different approximations and the convergence of many disciplines. Appropriate patient follow-up remains one of the most considerable difficulties. After the diagnosis, one or several antiseizure medications (ASMs) are prescribed to attain seizure freedom. In most cases, a specialist sees patients twice or thrice per year, relying on a few snapshots taken in the epilepsy monitoring unit (EMU) via video electroencephalography — in the best scenarios — or patient reports. However, it is known that the EMU creates an artificial environment where many seizure triggers are not present. Moreover, for several reasons, patients underreport their seizures, which creates a problem during the consultation for adjusting the treatments or in other scenarios like granting the possibility of driving. On the other hand, the autonomic nervous system (ANS) contributes to many clinical manifestations during seizures, sometimes being the predominant or sole manifestation. Seizure-mediated changes in the ANS have immediate consequences that vary from the need to change clothes due to urinary incontinence to an increased risk of dying. Even more, the action of the ANS has long-term consequences in people living with epilepsy (PWE), reflected in their increased cardiovascular risk. Technological advances have created the possibility of using devices for seizure detection outside the hospital. These include cameras, matrasses, and watches mainly used to detect generalized tonic-clonic seizures and warn caregivers when one occurs. However, other seizure types, such as focal impaired awareness seizures (FIAS) and typical absences, are of interest. Our research project used a wearable device (WD), the Byteflies Sensor Dot, in the EMU and at home; this dissertation summarizes the work. The Sensor Dot has the advantage that it is configurable to measure two-channel electroencephalogram (EEG) or electrocardiogram (ECG) and electromyography (EMG). We used the first configuration to measure brain activity with behind-the-ear electrodes and patches and the second to measure cardiac activity. The dissertation is divided into three parts. The first contains the introduction and research objectives, provides the base, and places the research project in specific domains. The second part consists of four chapters addressing each research project. Chapter 1 describes the experience of 16 patients with focal seizures who used the WD in the EMU during presurgical admission for a week and 16 outpatients for up to 8 months. Despite the patient's willingness to use the device at home, the likelihood of using it after six months decreased to 62%, mainly due to skin-related side effects. An automated EEG-based seizure detection algorithm developed by the engineers from the Department of Electrical Engineering at KU Leuven had a sensitivity of 52% in inpatients and 23% in outpatients. However, the sensitivity for extratemporal seizures was much lower. Chapter 2 describes our experience monitoring patients with absence seizures at home with wearable EEG (wEEG) to adjust ASMs. We followed 19 patients for up to 12 months, helping 79% of them to be absence-free on the last recording. Besides, we show that a machine learning algorithm helps reduce the time spent reviewing the wEEG to less than 5 minutes per 24 hours of recording. Usually, it takes between 20 and 30 minutes. In Chapter 3, we used a new approach to study cardiorespiratory interactions in focal epilepsy, using the wearable with the ECG configuration. This new approach separates the linear respiratory influences on the heart rate variability (HRV) and shows how respiration influences the periictal short-term HRV. Moreover, we used new methods to estimate respiratory sinus arrhythmia (RSA)—the variability in heart rate depending on the breathing frequency—on ultra-short-term ECG, finding a decrease in respiratory modulation of HRV during seizures. Chapter 4 represents an attempt to study sleep in the EMU with the wEEG and accelerometry using a deep-learning-based pipeline with a transfer learning approach. After further adaptations, this promising methodology could be used to study the bidirectional interaction between sleep and seizures at home during long-term monitoring. The third part comprises a general discussion, the recognition of other's contributions to this project, and a declaration of conflict of interests. The research project aimed to show different use cases for a wearable device in PWE. We found that detecting absence seizures at home is feasible and that ASMs can be adjusted accordingly. However, focal seizure detection requires further investigation using a multimodal approach. Moreover, besides patient monitoring, the information acquired can be used to better understand the basic mechanisms of the disease.