Prognosis of epilepsy using at-home EEG monitoring

Background Epilepsy is a serious neurological condition affecting over 600,000 people in the UK.

At present there are no robust markers of treatment response and clinical prognosis is effectively “watchful waiting”. Consequently, around 250,000 people have seizures that are poorly controlled by medication. We have developed a novel AI based biomarker of epilepsy: BioEP that can reveal seizure risk from brain data.

We have further evidence that the BioEP risk score changes with response to medication.

Aims and Objectives Our ultimate aim is a data acquisition and analysis pipeline, whereby an easy to wear headset will send electrical recordings of the brain wirelessly to a smartphone.

From these recordings – that will be collected in the home environment – a suite of AI algorithms will segment and extract relevant epochs of data, and perform analysis to provide prognostic markers.

Key objectives for this phase I project include: develop prototype app (including key modules for data handling, feature extraction and analysis), establish optimal number and location of EEG channels for headset, perform user testing, report project outcomes.

Project plan and methods used A lead engineer will work alongside our technical and scientific directors to create the app that will receive EEG recordings from an experimental headset designed by our development partner SensoMedical.

We will use a variety of AI algorithms including probabilistic PCA, random forests and SVM within key modules of the technology. Timelines for delivery We will complete the project in 12 months.

Anticipated Impact and Dissemination Alongside a prototype of our at-home monitoring technology, we will prepare a detailed product specification document and product development roadmap.

Working alongside people with epilepsy, we will disseminate our findings through patient networks (such as Epilepsy Action chapters), clinical epilepsy forums (such as the Birmingham Epilepsy Research Network), and where appropriate peer-reviewed publications.