Implementation of a Neonatal EEG Monitoring through Sonification on an Embedded Platform

The aim of this thesis is to implement a portable device able to put into practice a medical research that has being developed in University College of Cork. The idea is to realize an intuitive and persuasive solution for neonatal EEG monitoring systems used in healthcare facilities that enhances the classic approach by using sonification and deep learning AI, providing information about neonatal brain health to all neonatal healthcare professionals, particularly those without EEG interpretation expertise. As a matter of fact, interpreting the EEG activity of a newborn, requires very high knowledge and experience and it is still hard to detect an anomaly, such as a seizure or abnormal EEG background activity, since the symptoms are often contradictory. With the aid of an AI algorithm that has been already implemented, the percentage of detecting an abnormal brain functioning increases drastically and a contemporaneous real time sonification processing transforms the complex EEG signal into a sound much easy to interpret, minimizing the possibilities to make an error. The system has to be able to sample the EEG signals by using multiple channels, detect a possible anomaly, alert the medical staff, process the data in real time and stream video and audio on a device. The implementation presented in this thesis is mostly focused on real time signal processing and streaming, proposing a versatile and portable application that can be easily carried home and, in case of alert, can send data to a server in charge to process them and provide the streaming to the doctor that can decide if it is necessary to intervene immediately. The equipment chosen for this purpose are: • ADS1299 EEG (Texas Instrument): eight-channel, low-noise, 24-bit, simultaneous sampling delta-sigma analog-to-digital converters (ADCs) with a built-in pro-grammable gain amplifier (PGA), internal reference, and an onboard oscillator. The acquired samples are sent to a second platform through SPI. • Rasberry Pi model 3: it is in charge to received data through SPI and send them over a network to a Linux machine. • Laptop (Linux machine): this machine acts as a server that captures the data, applies the real time sonification algorithm and stream audio and video on a Python GUI. It is necessary to underline that this implementation represents the first demo of the full research project, indeed, many choices have been made for simplifying the initial work, such as the use of Raspberry Pi, and further modifications are contemplated to optimize the application and provide a system that is the best trade off in terms of power, cost and performances. The concept idea and demo will help the next researchers to have a starting point application that can perfectly work but it is still open to updates and refinements.