1) Localization of the Epileptogenic Zone

Over 20 million people in the world suffer from medically refractory epilepsy (MRE). MRE patients are frequently hospitalized and burdened by epilepsy-related disabilities, such as delay in neurocognitive development and incapacity in obtaining driving privileges - making this population a substantial contributor to the $16 billion dollars spent annually in the US treating epilepsy. Approximately 50% of MRE patients have focal MRE, where a hypothetical region in the brain, the epileptogenic zone (EZ), is the source of the seizures.

Treatment for MRE by resection of the EZ is often effective, provided the EZ is reliably identified. Focal epilepsy, however, is fundamentally a network-based disease. The EZ is composed and connected to a large-scale neuronal network whose other nodes may also exhibit abnormal neural activity either concurrently or subsequently. In patients without MRI detectable lesions, characterization and differentiation of the EZ from these other nodes in the epileptic network can be difficult, even with the use of invasive EEG recordings MRIs. Consequently, despite large brain regions being removed, guided by invasive evaluations, sustained surgical success rates barely reach 30%. Such disappointing outcomes associated with high morbidity rates are often due to imprecise and/or inaccurate localization of the EZ in association with attempting to treat a network disease using a single focal resection.

The goal of this research is to develop methods to create more precise and accurate maps of the EZ within epileptic networks in patients with MRE. These maps will strategically guide multiple laser ablations of network nodes resulting in the complete extinction of the epileptic activity, avoiding large resections and high morbidity rates. To achieve this goal, we are developing and testing computational tools that (i) estimate invasive EEG signals in “missing electrode” locations from measured EEG signals thus providing more dense coverage to clinicians, (ii) and converting EEG recordings to informative “maps” that show reliably where seizures start and spread. The maps will be created through the application of a novel algorithm that detects “fragility” of nodes in the epileptic network. The fragility of each node in a network (i.e., each implanted electrode) is defined as the minimum perturbation that can be applied to the electrode’s connectivity with its neighbors that destabilizes the entire network. Fragility is derived using dynamical systems theory and computed from electrophysiological data prior to seizure onset. In our preliminary study, the most fragile network nodes correspond to the EZ. We formed a team with expertise in epilepsy, electrophysiology, and dynamical systems to develop our tool.

 

This work is in collaboration with:

  • Jorge Gonzalez-Martinez (Cleveland Clinic)
  • Kareem Zaghloul (NIH)
  • Sara Inati (NIH)

This work is funded by:

  • NSF Graduate Research Fellowship
  • American Epilepsy Foundation Predoctoral Research Fellowship

 

 

2) Closed-loop Stimulation to Suppress Seizure Genesis in Epilepsy

Epilepsy affects approximately 70 million people worldwide. About 30% of epilepsy patients are drug resistant and must consider invasive alternatives such as resective surgery, and electrical stimulation therapy. Surgical candidates must have a well-localized focus in an area outside of eloquent brain structures. Although surgery can dramatically improve the lives of patients, it is irreversible and outcomes are highly variable (30-70% success rates). Electrical stimulation, on the other hand, is reversible and has great potential. Chronic open-loop stimulation has shown some efficacy but does not account for dynamic brain activity and the continuously changing state of the patient, making it suboptimal and crude. To maximize therapeutic effects, new methods must be developed for fine dynamic tuning of stimulation parameters in a patient-specific manner. Closed-loop therapy provides an attractive option that minimizes intervention by limiting the delivery of therapy to times when the patient is in need.

Efforts have been made to develop “closed-loop” stimulation strategies using different protocols, yet none provide a highly effective and reliable solution. Most closed-loop strategies proposed and studied are “responsive switches” and often are “on” the majority of the time. These strategies wait until a seizure is detected (via a detection algorithm) and then stimulated with a fixed pattern to suppress the seizure. In contrast, we are implementing a real-time closed-loop control system that continuously steers the neural network away from seizure genesis entirely using adaptive stimulation patterns that change with EEG measurements - avoiding seizure detection and seizures altogether.

To meet this objective, are using in vivo experimental data to develop an innovative mathematical model that characterizes fundamental neural dynamics during seizure genesis, and the effects of different electrical stimuli on neural activity leading to seizure genesis. Based on this model, we will then design and implement a feedback controller that monitors neural activity in real-time to prevent seizures from evolving in the network. In particular, the controller will steer temporal patterns of stimulation to disrupt pre-seizure activity with minimal energy consumption. To accomplish our goals, we have assembled a highly interdisciplinary team with expertise in system identification, control, and experimental neurophysiology.

 

This work is in collaboration with:

  • Noah Cowan (JHU)
  • Yitzhak Schiller (Technion)

This work is funded by:

  • NIH NINDS R21
  • HHMI Gilliam Fellowship