Medical Innovation Exchange

The role of AI in advancing neuromodulation treatment for epilepsy

Epilepsy is one of the most common neurological diseases, affecting around 50 million people worldwide, and carries a debilitating burden for the patient. It is characterized by recurrent seizures which occur as a result of excessive electrical discharge in a group of brain cells and are extremely heterogeneous.

Seizures vary in their symptoms, severity and frequency both between patients and between an individual patient’s seizure events in terms of; the brain region affected, physiological response, neurological response, physical response, and triggers. As a result, seizures are notoriously difficult to detect or predict, and the efficacy of treatments is variable, which severely impacts an individual’s social freedom and employability. A single seizure, for instance, will bar someone from driving for 6 months, and those with recurrent seizures must be seizure-free for 5 years before they are allowed to drive.

It’s no surprise, then, that people living with epilepsy report the two main causes of their daily challenges to be the seizure unpredictability and the inefficacy of their medications.

Antiepileptic drugs (AEDs) are the most common treatment option, helping to suppress seizures in around 60-70% of patients, however 30-40% of the patient population are unresponsive to AEDs and are therefore considered to be refractory.

For refractory epilepsy patients, limited treatment options exist. Brain surgery may be considered for those eligible, which in the UK is only 1.5% of epilepsy patients, but for most, the remaining option is neuromodulation therapy.

The Current State of Neuromodulation for Epilepsy

Neuromodulation involves the alteration of nerve activity through targeted electrical stimulation and the two specific forms involved in the treatment of epilepsy are deep brain stimulation (DBS) and vagus nerve stimulation (VNS).

DBS involves the surgical implantation of a neurostimulator device, usually in the chest, and electrodes, which are connected directly to various parts of the brain to deliver precise electrical stimulation that engenders symptom suppression. One study of 81 participants found that patients experience a higher reduction in seizure frequency the longer a DBS system has been fitted, from a median decrease of 41% in the first year, to 68% in the third. However, DBS is still considered a relatively new treatment so the jury is still out on its efficacy, and there are some serious risks associated with it, including brain hemorrhage, infection, depression and memory loss.

VNS, on the other hand, has been used in the treatment of epilepsy since the first implantable device was approved by the FDA in 1997. Invasive VNS involves the surgical implantation of a neurostimulator, usually in the chest, and electrodes, which are connected to the vagus nerve in the neck; the longest and most widely distributed of the 12 cranial nerves in the human body that carries motor, sensory and parasympathetic information. Its safety and efficacy has been established in multiple clinical studies with more than a 50% average reduction in seizure frequency reported in the first year by 20-40% of patients. Despite this, adverse events are still associated with invasive VNS, from vocal cord paresis and lower facial weakness, to infection, bradycardia and asystole. Any neurosurgeon will also testify to their issues surrounding lead breakage and battery life. For these reasons, non-invasive VNS (nVNS) devices have been developed, such as gammaCore™.

But this is not the most exciting innovation occurring in neuromodulation treatment for epilepsy. It is instead the dawn of a closed-loop approach whereby electrical stimulation is automatically informed by measurement of real-time physiological changes relating to seizures, without any action required by the patient. Previously, stimulation has been delivered either manually via a portable magnet or at set regular intervals.

In DBS, approved devices include Medtronic’s Percept PC and NeuroPace’s RNS System, which record internal brain waves (local field potentials) and surface brain waves (via electrocorticography or ECoG), respectively, to inform personalized stimulation protocols and enable more proactive seizure detection and treatment.

With VNS, however, the opportunity exists to capture multiple parameters which can be used to both predict seizures and inform automatic stimulation. This is due to the extensive reach and afference of the vagus nerve, which allows it to modulate the function of higher brain centers, and underlies its potential use in treating many disorders (e.g. epilepsy, depression, rheumatoid arthritis, Crohn’s disease, migraine, obesity, diabetes, heart failure, tinnitus, stroke). The first steps have already been taken by LivaNova’s AspireSR device. This device senses the individual’s heart rate via electrocardiography (ECG), meaning in individuals whose seizures are initiated by a substantial increase in heart rate, automatic VNS can be triggered to stop or shorten a seizure. One study of 44 patients found that 71% reported ?50% reduction in seizure burden after the AspireSR device was inserted.

Notably, however, only 57% of seizures cause an increased heart rate. The roadmap for future exploration into closed-loop VNS for epilepsy should therefore be more extensive investigation into the feasibility of measuring parameters across brain (EEG), nervous (EDA), cardiac (ECG) and respiratory (PPG) function to detect and predict seizures. A device that could continuously monitor some or all of these functions holds the potential to not only automate but also personalize care on a hitherto unimaginable scale.

The Dawn of AI & Intelligent, Closed-loop Neuromodulation

And that’s where AI and machine learning have a massive role to play. Until now we have not had access to the powerful algorithms needed to ingest, analyze and instrumentalize huge volumes of real-time physiological data, but advancements in recent years mean that intelligent neuromodulation systems are possible.

Not only do they have the potential to inform personalized therapeutic interventions in real time based on multiple physiological parameters, but also, over time, to map the unique physiology and disease state of the individual, learn their baseline and become increasingly effective at spotting deterioration from this baseline to inform intervention. In short, as they become more and more tailored to the patient, they become more effective.

There is the opportunity here to not only transform our approach to treating disease and empower epilepsy patients (and beyond) with improved outcomes and quality of life, but also to expand our understanding of disease mechanisms through the mapping of neural circuits and the exact effect of electrically induced modulation on the disease state.

Epilepsy is the perfect starting point for intelligent, closed-loop neuromodulation. VNS devices still have a long way to go, but the therapeutic potential of the vagus nerve, the advent of AI and machine learning in healthcare and the opportunity to establish truly personalized digital medicine are enough to warrant further exploration into VNS.

Photo: metamorworks, Getty Images

Bipin Patel

Leave a Comment

Your email address will not be published. Required fields are marked *

Leave the field below empty!