Inspiration

EpiPredictor, is a first ever idea to integrate mobile and handset technology (muse) to define totally new way to Predict Epileptic Seizures before they take place. We alarm, Patients / Doctors / Loved ones through synchronized web and mobile applications all at once, so to take care and prevent any mishap.

EpiPredictor aim is to improve the life of people who are suffering with seizures and allowing them to live more freely.

Per the Institute of Medicine report 'Epilepsy Across the Spectrum,' In the United States alone, epilepsy affects 3.2 million people with 150,000 new epilepsy cases every year. Of these, as many as 40% of patients with epilepsy are misdiagnosed and suffer from uncontrolled seizures, which limits their independence at best and can result in death at worst. People often suffer for years before their epilepsy can be correctly identified and treated. The monetary cost of epilepsy misdiagnosis and mismanagement of healthcare is $12.5 billion annually.

A teammate’s brother has epilepsy with frequent seizures. He may not always be in a safe spot when a seizure hits. This system will empower him by helping him get to a safe spot before the seizure and will alert his family. This system will help him do activities and apply for jobs that he may not have been able to because they required complete attention(Driving). Dogs have been used to predict seizures but they are not accessible to all patients since the dogs require a lot of training/expense, are not available in all countries, or the patient may be allergic. This system would hopefully be accessible to many more.

What it does

The application predicts seizures in real-time and notifies the patient and the patient's family/friends that (s)he is going to have a seizure based on his/her brain activity. The system also takes into account the patient’s movement and heartbeat and saves all data for review by their Doctor.

How we built it

We used a Muse headband to measure the persons brainwaves. We made an android application to gather the sensor data from the headband over bluetooth. The app will talk to a server over a REST API. We made a server and leveraged InfluxDB to store the data being sent to the server from the android app. We used a deep neural network model to process that data and give a prediction which is sent back to the android app. The app itself will send texts/emails to family/friends and log all the data.

Challenges we ran into

There are many deep neural networks with their own different benefits. Choosing one that would do what was needed and had enough surrounding documentation for efficient implementation was challenging.

For the demo we can’t actually test for a seizure, so we decided to test the neural network with data on focus/attention/relaxation. Everything except the data should be the same, so it should help prove that the muse headband and the network together are able to parse brainwaves accurately enough to be useful.

A lot of epilepsy data that could be used to train the network is behind a paywall. Finding good data and figuring out how it has been formatted took a lot of time.

The Muse headband is very accurate but can be dislodged easily as it was originally designed for still meditation. A new headband design with a snug fit for everyday use would be something to consider in the future.

The muse headband has a great API but not as much explanation for how each function may be used or what exactly it will return. It resulted in a lot of trial and error.

The fitbit was intended to monitor heart-rate and seizure bodily motion but their API and data management is not very conducive to sharing with a third party application on short notice like this project.

There is not enough time to implement full HIPAA medical data compliance.

Accomplishments that we're proud of

The android app and database system were up and running well. The research surrounding networks took awhile but we learned a lot as team.

What we learned

We learned about different types of deep neural networks, about android studio, brainwaves, EEG and epilepsy wave signatures, and database systems.

What's next for EpiPredictor

Adding HIPAA compliance, more development on the neural network, increased datasets, remade muse headband, cross-platform, coming to your applewatch or fitbit!

Built With

Share this project:
×

Updates