Frequently Asked Questions

Actually, no! BioEP is not a traditional “black box” AI system.

BioEP focuses on analysing the background activity of the brain, which often appears normal during clinical review of routine EEG. However, within the signal there is vital information about how the brain is behaving and how prone it is to generating seizures. BioEP uses a number of concepts from mathematics to reveal hidden insights from EEG that might otherwise go unnoticed. This approach is quite different to a traditional AI model, which might be given a set of defined features (such as interictal spikes or discharges) and attempt to focus on detecting clear-cut features like interictal discharges (IEDs). BioEP adopts a different strategy—using pure mathematics to reveal hidden insights that might otherwise go unnoticed.

Let’s break it down further: BioEP uses a suite of algorithms to study the EEG data that has been recorded as part of standard clinical workup. Each algorithm evaluates a specific property of the EEG that has been shown in peer-reviewed literature to be altered in people with epilepsy. We term these digital biomarkers. Similar to physiological biomarkers, like those that can be calculated from blood, the process underpinning each digital biomarker that informs our BioEP rating is fully transparent. The key properties we examine include:

• EEG signal properties: BioEP calculates fundamental aspects of the brain’s electrical activity, such as the frequency and amplitude of the EEG signal, which help characterize brain function. This approach is built upon Fourier analysis and signal processing.

• Brain network properties: The brain functions as a network of interconnected regions that constantly communicate with each other. BioEP uses concepts from graph theory to assess the structure and strength of these connections, providing an indication of how different brain regions interact.

• Brain dynamics: Our understanding of how seizures emerge in the brain from apparently normal brain states has grown rapidly in recent years. It is now understood that the interplay between brain networks and the dynamics they support are critical ingredients for seizures to occur. To assess this BioEP uses concepts from mathematical modelling. Essentially, we construct a computer representation of brain networks, informed from the networks inferred from the collected EEG. The computer model is a set of differential equations, these describe how brain dynamics vary over time. Simulations of these models under a variety of conditions rapidly reveal the ease with which seizures can emerge. Parameters that characterize this ease from a modelling perspective provide the final biomarkers used to inform the BioEP rating.

Once these biomarkers are calculated, BioEP uses a supervised statistical model to classify the EEG from very supportive to very unsupportive of an epilepsy diagnosis. This model has been trained on hundreds of inconclusive EEGs from people with epilepsy and with other common conditions (such as non-epileptic attack disorder) so that we understand how the biomarkers we use appear in people with epilepsy and in people whose symptoms are caused by conditions other than epilepsy.

This approach combines well-understood scientific principles with the power of machine learning, offering a transparent, deterministic, and interpretable method to assist clinicians in their decision-making.

BioEP is designed to analyse EEGs where no interictal epileptiform discharges (IEDs) are present, meaning there is insufficient diagnostic information in the EEG for clinicians to determine seizure susceptibility based on visual analysis alone. In such instances, BioEP provides valuable complementary information, enhancing clinicians’ diagnostic decision-making. 

Currently, BioEP has been tested and validated for adult (18+) patients with various types of epilepsy, including generalised and focal (aware and unaware) epilepsy. In the near future, we aim to expand validation efforts to include paediatric populations. Additionally, we are conducting a prospective observational study to investigate the association between BioEP and self-reported seizure control during ASM treatment.

BioEP processes identifiable patient data, but only to the absolute minimum necessary to ensure that reports are clinically useful for healthcare professionals. Access to this data is strictly limited to authorized individuals and all processing is conducted in full compliance with NHS data protection regulations.

If identifiable patient data is used beyond clinical reporting—such as for product improvement or clinical research—patients will receive clear, detailed information and must provide explicit consent to participate before their data is used.

We work closely with clinics to ensure that all data protection measures align with privacy regulations and local policies.  Data security is our top priority.

BioEP is designed to work with as little as 5 minutes of EEG recording, making it highly accessible for clinical use. The EEG should include a minimum of 19 channels, following the international 10-20 electrode placement system, or a montage from which re-referencing can result in a 10-20 representation. Importantly, hyperventilation and photic stimulation are not required for BioEP to function, however if such activation methods are utilised in the EEG this will not impact the ability of BioEP to provide a rating of support for an epilepsy diagnosis.

The EEG data should be uploaded in EDF format. Once data is uploaded in this format, the rest of the process is fully automated, ensuring quick and easy integration with your clinical workflow.   

BioEP generates a detailed PDF report within approximately 10 minutes after an EEG is uploaded to the Neuronostics Platform. The processing time may vary slightly depending on the length of the EEG recording and the number of requests for BioEP currently being made.

It is important to note that the digital biomarkers that BioEP uses have been shown to vary in awake EEG. Therefore, BioEP is not suitable for use with purely sleep EEGs. Further, while longer recordings can be uploaded, such as 24-hour studies, BioEP will only utilise periods of awake, resting, data.

When BioEP receives an EEG, it divides the recording into smaller segments called epochs. Each epoch is then assessed for artifacts. If e.g. large fluctuations are detected—such as those caused by movement or eye-blinking—the affected epoch is effectively excluded from the analysis. Additionally, the biomarkers are calculated across multiple epochs, and the final value for each biomarker is determined as the median of these segments. This method helps to minimize the impact of outliers and maintain the accuracy of the analysis. Note that segments including hyperventilation and photic stimulation will also be treated as artefacts, as BioEP is designed to analyse resting EEG.

BioEP is a Class 1 Medical Device under 93/42/EEC Medical Device Directive (MDD) and is UKCA marked. Additionally, we are currently in the process of obtaining FDA approval to expand our regulatory compliance in the United States.

Neuronostics are ISO 13485 certified, Cyber Essentials certified, and compliant with DTAC and NHS DSPT.

BioEP is designed for seamless integration into daily clinical practice, requiring minimal training to use effectively. We offer a brief training session to address any initial questions and provide an overview of the tool’s functionality. After the initial setup, our team remains available for ongoing support and technical assistance to ensure smooth usage and address any queries that may arise. Additionally, the BioEP platform includes a comprehensive user manual with step-by-step instructions, accessible from the home screen on the BioEP platform for easy reference.