Seizure Sensor

Fall 2024

Project developed for SteelHacks XI, a 24-hour hackathon run by the Pitt Computer Science Club.

This project won Best Healthcare Hack and Second Place Overall.

Inspiration

The inspiration for Seizure Sensor came from a deeply personal place. Kevork's brother, who suffers from epilepsy, faces the constant threat of life-threatening seizures during the night. His parents relied on a basic webcam system to monitor him, often missing critical moments when seizures occurred. This pressing need for a more reliable monitoring system drove us to create Seizure Sensor.

What does it do?

Seizure Sensor is an innovative monitoring system that combines:

  • Wearable technology with EMG sensors, accelerometers, and heart rate monitors

  • Computer vision using OpenCV for motion detection

  • A central MQTT server on a Raspberry Pi for data integration

  • An ESP32-powered visual and auditory alarm system

This integrated system provides accurate, timely detection of seizure activity and alerts caregivers promptly, significantly improving the monitoring process for people with epilepsy.

How did we build it?

  1. Initially, we attempted to develop an iOS app using Apple Watch for biometric data.

  2. Due to challenges with the Apple development platform, we pivoted to creating our own wearable using ESP32 microcontrollers.

  3. We developed custom hardware incorporating various sensors for comprehensive biometric data collection.

  4. We set up an MQTT server on a Raspberry Pi to integrate and process all sensor inputs.

  5. We implemented computer vision capabilities using OpenCV for motion detection.

  6. Finally, we created an alert system using ESP32 to provide clear visual and auditory alarms when a seizure is detected.

Kushal and me, troubleshooting our program for the wearable
Kevork and Kushal, troubleshooting hardware integration
Kevork and Oday, working on the MQTT server
Testing the heart rate sensor and Myoware muscle senor running on ESP32

Challenges We Encountered

  1. Difficulties interfacing with the Apple development platform, leading to our pivot to custom hardware.

  2. Integrating multiple sensory inputs (EMG, accelerometer, heart rate) into a cohesive system.

  3. Implementing effective computer vision algorithms for seizure detection.

  4. Ensuring the system's reliability and accuracy without inducing actual seizures for testing.

What We Learned

  1. The importance of flexibility in the development process, as demonstrated by our successful pivot from iOS to custom hardware.

  2. Techniques for integrating multiple sensory readings into an MQTT server for accurate biometric monitoring.

  3. The complexities and potential of computer vision in medical applications.

  4. The critical balance between innovation and ethical considerations in medical technology development.

  1. Further refine and enhance the computer vision model with specialized training on seizure footage.

  2. Conduct more extensive testing to validate the system's efficacy and reliability.

  3. Explore partnerships with medical institutions for clinical trials.

  4. Investigate the potential for miniaturization and commercialization of the technology.

  5. Develop a user-friendly interface for caregivers to easily monitor and respond to alerts.

Next Steps

Best Healthcare Hack Award

Second Place Overall Award

Team Members: Oday Abushaban, Tyler Hansen, Kushal Parekh, and Kevork Zeibari

Next

Project HEARTH