EpihunterDetects, logs and signals
silent, difficult to notice
epileptic absence seizures.

Explore the case

About the client

Epihunter is a Belgian startup building digital solutions to help normalize the daily life of people with epilepsy. They detect, log and signals silent, difficult to notice epileptic absence seizures. Epihunter simply turns on a light when the brain switches off.


The problem

Absence seizures are the silent, often unnoticed epilepsy seizures which pauses the person’s brain and causes the affected person to not register anything in their vicinity for a short period heime. These seizures occur at random and their frequency can vary significantly from a couple of seizures per day to a couple of seizures per month. They have a tremendous impact on the quality of life of 150,000 young children. Unawareness of the person’s environment results in low self-esteem and frustration.

Epihunter's solution?

Using electroencephalogram (EEG) data, it is possible to record and detect these seizures. In combination with a headset, Epihunter developed an Android application capable of monitoring live EEG-data and showing feedback if a seizure was detected.

All recorded data gets sent to an online web platform, which the users can access using a dashboard.

The solution

The solution Sentigrate worked out consisted of 3 steps:

  1. Seizure detection:
    • Labeling data with a self-developed annotation tool
    • Data preprocessing and handling different kinds of noise
    • Developing a suitable machine learning model with Tensorflow
    • Evaluating relevant metrics and monitoring the selected model’s results
  2. Running the algorithm: in order to detect the seizures, the algorithm should process the data and detect the seizures in real-time. Therefore Sentigrate co-developed an Android application capable of running the Tensforflow model and providing appropriate audiovisual feedback to the user.
  3. Data visualization: an online secure dashboard was created allowing users to gain insight in their seizure frequency and at what time seizures typically occur. The dashboard was developed with Angular 4+.


increase in sensitivity


increase in specificity

Increase in sensitivity and specificity

The machine learning model was verified on both clinical data and real life data collected by Epihunter.

Both achieved a sensitivity (true positive rate) of more than 90%, meaning on average nine in ten seizures are correctly detected. But also the specificity (true negative rate) is above 90%, meaning a detection is right in nine out of ten cases.

Epihunter explanatory video

Company Facts

Number and facts about the case

Output of data via app
Epihunter Companion