Presented By: Biomedical Engineering
PhD Defense: Jared Scott
Temporal Characteristics of High-frequency Oscillations as a Biomarker of Human Epilepsy
NOTICE: This event will be held via Zoom. The link will be placed below.
https://umich-health.zoom.us/j/97604985906?pwd=N1Y1UXEvNXMxdjlnVkpjUFZHQkRhdz09
Epilepsy is a debilitating neurological disorder characterized by recurrent spontaneous seizures. While seizures themselves adversely affect physiological function for short time periods relative to normal brain states, their cumulative impact can significantly decrease patient quality of life in myriad ways. For many, anti-epileptic drugs are effective first-line therapies. One third of all patients do not respond to chemical intervention, however, and require invasive resective surgery to remove epileptic tissue. While this is still the most effective last-line treatment, many patients with ‘refractory’ epilepsy still experience seizures afterward, while some are not even surgical candidates. Thus, a significant portion of patients lack further recourse to manage their seizures – which additionally impacts their quality of life.
High-frequency oscillations (HFOs) are a recently discovered electrical biomarker with significant clinical potential in refractory human epilepsy. As a spatial biomarker, HFOs occur more frequently in epileptic tissue, and surgical removal of areas with high HFO rates can result in improved outcomes. There is also limited preliminary evidence that HFOs change prior to seizures, though it is currently unknown if HFOs function as temporal biomarkers of epilepsy and imminent seizure onset. No such temporal biomarker has ever been identified, though if it were to exist, it could be exploited in online seizure prediction algorithms. If these algorithms were clinically implemented in implantable neuromodulatory devices, improvements to quality of life for refractory epilepsy patients might be possible. Thus, the overall aim of this work is to investigate HFOs as potential temporal biomarkers of seizures and epilepsy, and further to determine whether their time-varying properties can be exploited in seizure prediction.
In the first study we explore population-level evidence for the existence of this temporal effect in a large clinical cohort with refractory epilepsy. Using sophisticated automated HFO detection and big-data processing techniques, a continuous measure of HFO rates was developed to explore gradual changes in HFO rates prior to seizures, which were analyzed in aggregate to assess their stereotypical response. These methods resulted in the identification of a subset of patients in whom HFOs from epileptic tissue gradually increased before seizures.
In the second study, we use machine learning techniques to investigate temporal changes in HFO rates within individuals, and to assess their potential usefulness in patient-specific seizure prediction. Here, we identified a subset of patients whose predictive models sufficiently differentiated the preictal (before seizure) state better than random chance.
In the third study, we extend our prediction framework to include the signal properties of HFOs. We explore their ability to improve the identification of preictal periods, and additionally translate their predictive models into a proof-of-concept seizure warning system. For some patients, positive results from this demonstration show that seizure prediction using HFOs could be possible.
These studies overall provide convincing evidence that HFOs can change in measurable ways prior to seizure start. While this effect was not significant in some individuals, for many it enabled seizures to be predicted above random chance. Due to data limitations in overall recording duration and number of seizures captured, these findings require further validation with much larger high-density intracranial EEG datasets. Still, they provide a preliminary framework for the eventual use of HFOs in patient-specific seizure prediction with the potential to improve the lives of those with refractory epilepsy.
Chair: Dr. William Stacey
https://umich-health.zoom.us/j/97604985906?pwd=N1Y1UXEvNXMxdjlnVkpjUFZHQkRhdz09
Epilepsy is a debilitating neurological disorder characterized by recurrent spontaneous seizures. While seizures themselves adversely affect physiological function for short time periods relative to normal brain states, their cumulative impact can significantly decrease patient quality of life in myriad ways. For many, anti-epileptic drugs are effective first-line therapies. One third of all patients do not respond to chemical intervention, however, and require invasive resective surgery to remove epileptic tissue. While this is still the most effective last-line treatment, many patients with ‘refractory’ epilepsy still experience seizures afterward, while some are not even surgical candidates. Thus, a significant portion of patients lack further recourse to manage their seizures – which additionally impacts their quality of life.
High-frequency oscillations (HFOs) are a recently discovered electrical biomarker with significant clinical potential in refractory human epilepsy. As a spatial biomarker, HFOs occur more frequently in epileptic tissue, and surgical removal of areas with high HFO rates can result in improved outcomes. There is also limited preliminary evidence that HFOs change prior to seizures, though it is currently unknown if HFOs function as temporal biomarkers of epilepsy and imminent seizure onset. No such temporal biomarker has ever been identified, though if it were to exist, it could be exploited in online seizure prediction algorithms. If these algorithms were clinically implemented in implantable neuromodulatory devices, improvements to quality of life for refractory epilepsy patients might be possible. Thus, the overall aim of this work is to investigate HFOs as potential temporal biomarkers of seizures and epilepsy, and further to determine whether their time-varying properties can be exploited in seizure prediction.
In the first study we explore population-level evidence for the existence of this temporal effect in a large clinical cohort with refractory epilepsy. Using sophisticated automated HFO detection and big-data processing techniques, a continuous measure of HFO rates was developed to explore gradual changes in HFO rates prior to seizures, which were analyzed in aggregate to assess their stereotypical response. These methods resulted in the identification of a subset of patients in whom HFOs from epileptic tissue gradually increased before seizures.
In the second study, we use machine learning techniques to investigate temporal changes in HFO rates within individuals, and to assess their potential usefulness in patient-specific seizure prediction. Here, we identified a subset of patients whose predictive models sufficiently differentiated the preictal (before seizure) state better than random chance.
In the third study, we extend our prediction framework to include the signal properties of HFOs. We explore their ability to improve the identification of preictal periods, and additionally translate their predictive models into a proof-of-concept seizure warning system. For some patients, positive results from this demonstration show that seizure prediction using HFOs could be possible.
These studies overall provide convincing evidence that HFOs can change in measurable ways prior to seizure start. While this effect was not significant in some individuals, for many it enabled seizures to be predicted above random chance. Due to data limitations in overall recording duration and number of seizures captured, these findings require further validation with much larger high-density intracranial EEG datasets. Still, they provide a preliminary framework for the eventual use of HFOs in patient-specific seizure prediction with the potential to improve the lives of those with refractory epilepsy.
Chair: Dr. William Stacey
Related Links
Explore Similar Events
-
Loading Similar Events...
