We've created a wireless EEG processing system which uses 6 electrodes on the scalp, close to the hairline, and a feature extraction algorithm to detect the onset of a seizure. This algorithm then sends an alert to the patient asking if they need medical attention. It immediately begins recording EEG data upon seizure detection for future processing and research. The device can alert an emergency contact, such as the patient's parent or caretaker. The real-time data can also be monitored through a bluetooth connection on computers and mobile devices.
Seizure activity detection is performed with a line length threshold algorithm running on each of the 6 channels. Line length was determined to be the best feature to extract from the signal based on computational complexity and reliability. Attached is an example of actual EEG data from an epilepsy patient, where blue represents raw data, green the extracted feature (Line Length), and red the detection of seizure activity.
This entire platform is currently wearable and has the potential to be shrunk down even further with PSoC4 devices. The current prototype requires multiple microcontrollers, but these can later be merged into a single processor that supports an RTOS and WiFi.