Decoding Seizure Activity

Challenge ID


Champion Name

Pierre Wijdenne

Project Track


Team Members

Yuvra O'Neill, Kiefer Peck

Skills Needed

Machine Learning, Data Science, Matlab

Challenge Description

Decoding Seizure Activity

Our understanding of brain function under normal and various pathological conditions remains limited, due in large part to our inability to monitor subtle electrical signals from networks of interconnected neurons. Thus, the underlying causes of many neurological disorders like epilepsy remain unknown, thereby precluding our ability to repair or compensate the perturbed brain function.

In turn, this lack of fundamental knowledge stems from our inability to monitor neural activities from complex, synaptically connected networks at a high enough spatio-temporal resolution and over extended time periods. In short, our ability to improve the quality of life for people with neurological disorders is limited by the quality of the data we can collect. In an attempt to fill this gap, we prototyped and developed several microelectrodes (biosensors) that allow higher neural electrical signal recording and that are now routinely interfaced with a variety of homogeneous cell culture preparations maintained in controlled environments in-vitro. In parallel, we also developed a strong network of electrophysiologists, neurosurgeons and micro/nano fabrication engineers that are guiding us during the technological development and who expressed deep interest into the potential of our work.

Thanks to the initial in-vitro recordings we collected, we believe that it is possible to detect and predict seizures a few seconds before they happen, thus generating the ability to alert patients to avoid a potentially risky situation (cooking, climbing stairs, etc.). While other companies have so far been unsuccessful in predicting seizures in patients, mainly due to poor signal detection and analysis, our 3d-electrodes could, in theory, offer a high enough resolution and therefore permit a better analysis of the signals. The challenge we are submitting is to develop a machine learning algorithm that would predict these seizures. We are offering to share the data that we collected (brain activity recordings) and work with innovators to decode these pre-seizure activities.