Eline Schaft

Approximately 50 million people worldwide have epilepsy. Epilepsy is caused by a disturbance of electrical activity in the brain, which can cause seizures. For a proportion of people with epilepsy, medication does not provide sufficient seizure control. For these patients, epilepsy surgery can be a potentially curative treatment. The goal of epilepsy surgery is to achieve lifelong seizure freedom. Currently, the reported proportion of patients who become seizure free after surgery ranges between 36 and 76%, indicating that there is room for improvement.

In epilepsy surgery, it is essential to completely remove the epileptogenic zone: the brain region responsible for seizure generation. Accurately localizing this zone is challenging and is performed prior to surgery using various diagnostic investigations, such as EEG, MRI, and sometimes additional advanced imaging techniques. Based on these investigations, a surgical plan is formulated. During surgery, this plan can be evaluated and, if necessary, adjusted using intraoperative electrocorticography (ioECoG), in which brain activity is recorded directly from the cortical surface. These signals are assessed in real time during surgery. Traditionally, attention is focused on two types of abnormal signals, referred to as biomarkers: interictal spikes and high-frequency oscillations (HFOs). Interictal spikes are brief, sudden discharges of neuronal activity that indicate increased excitability of brain tissue. HFOs are very fast rhythmic oscillations at high frequencies (above 80Hz), which can be subdivided into ripples (80-250Hz) and fast ripples (250-500Hz). Nevertheless, there is ongoing debate regarding the use and clinical relevance of spikes and HFOs. This debate concerns both which biomarker is most informative for different patient groups and practical questions related to how these signals should be recorded and interpreted.

This thesis investigates the potential of ioECoG beyond traditional visual assessment and aims to contribute to a more accurate, efficient, and automated delineation of the epileptogenic zone. The work focuses on two complementary themes: advanced signal analysis of ioECoG and the interpretation of ioECoG, including the clinical implementation of high-resolution grids.

In Chapter 2, we investigated the temporal relationship between interictal spikes and HFOs when they occur simultaneously. By studying only patients who became seizure free after surgery, the resected tissue could be considered representative of the tissue responsible for seizure generation. We found that concurrently occurring interictal spikes and ripples were more frequently present in resected, epileptic tissue. In addition, subtle temporal features of these signals, such as their timing relative to one another, were found to provide information about which brain tissue was epileptic.

In Chapter 3 and Chapter 4, we examined whether specific signal characteristics of the ioECoG could support in identifying epileptic tissue. In Chapter 3 we investigated whether variability in the complexity of the ioECoG signal, calculated as spectral entropy, was associated with surgical outcome. We found that this measure contains information both for distinguishing epileptic from healthy tissue and for estimating the likelihood of postoperative seizure freedom. In Chapter 4, artificial intelligence was used to investigate which frequency components of the ioECoG contribute to distinguishing epileptic from non-epileptic tissue. We found that certain changes in background activity, which are not readily visible to the naked eye, can nonetheless be informative. Whereas Chapter 2 and Chapter 4 focused on ioECoG recordings obtained prior to resection and included only patients who were seizure free, Chapter 3 included ioECoG recordings acquired both before and after resection in patients with favorable as well as unfavorable surgical outcomes.

In Chapter 5, we investigated whether the way ioECoG signals are displayed influences the detection of interictal spikes and HFOs. We compared different visual representations of the same ioECoG recordings and assessed how well the various biomarkers could be identified. We found that the choice of montage has a major impact on what is observed, and that certain montages are more suitable than others for detecting specific biomarkers. This highlights that the interpretation of ioECoG depends not only on the signal itself, but also on how it is presented.

In Chapter 6, we examined the relationship between the location of abnormalities visible on MRI and the location of biomarkers detected with ioECoG. We found that this relationship differs depending on the underlying pathology, emphasizing the importance of a personalized interpretation of ioECoG. In Chapter 7, we describe the clinical implementation of high-resolution grids for ioECoG recordings. These grids contain four times as many electrodes as conventional low-resolution grids. Their use allows for the detection of a greater number of biomarkers, particularly fast ripples, and enables detailed visualization of the spatial spread of interictal spikes.

The results of this thesis demonstrate that ioECoG can provide valuable information during epilepsy surgery, but that no single biomarker or method is sufficient on its own to reliably identify epileptic tissue. The significance of interictal spikes, HFOs, and more subtle signal characteristics varies between patients and underlying pathologies. In addition, the interpretation of ioECoG is influenced by technical choices, such as recording and display methods, as well as by practical factors in the operating room. By combining advanced signal analysis with clinical expertise and by carefully implementing new recording techniques, such as high-resolution grids, ioECoG can evolve into a more objective and reproducible tool. These developments offer perspective for further improving epilepsy surgery, with the ultimate goal of increasing the likelihood of seizure freedom.