By 
Zurich researchers have created a small, efficient device to decode brainwaves using artificial neurons. The chip analyzes brainwaves from epilepsy patients to determine which brain regions are responsible for epileptic seizures. This provides new possibilities for treatment.
The current neural network algorithms achieve impressive results and solve many problems. These algorithms require too much processing power for the electronic devices that run them. Artificial intelligence (AI) systems cannot compare to an actual brain processing sensory information and interacting with the environment in real time.
High-frequency oscillations detected by Neuromorphic chip
Neuromorphic engineering bridges the gap between artificial intelligence and natural intelligence. This approach was used by an interdisciplinary research team from the University of Zurich and ETH Zurich to create a chip that recognizes complex biosignals reliably and accurately. This technology was able to detect HFOs, and previously recorded high-frequency oscillations (HFOs), and the scientists were competent to use it to do so. These waves were measured with an intracranial electroencephalogram (EEG) and be promising biomarkers in identifying brain tissue responsible for epileptic seizures.
Complex, compact, and efficient in energy use
Researchers first created an algorithm to detect HFOs using the brain’s natural neural networks. This tiny network is called the “spiking neural network” (SNN). The second step was to implement the SNN in a small piece of hardware, which receives neural signals via electrodes. This is much more energy-efficient than conventional computers. This allows calculations to be performed with high temporal resolution without needing the internet or cloud computing. Giacomo Inveri, professor at UZH’s Institute for Neuroinformatics and ETH Zurich, says, “our design allows us to recognize a spatiotemporal pattern in biological signals in actual time.”
Measuring HFOs outside and in operating rooms
Researchers are now working to develop an electronic system that can reliably detect and monitor HFOs in real time. The system can be used in neurosurgical operations as an additional diagnostic tool.
HFO recognition isn’t the only area where HFO recognition could be helpful. The team’s long-term goals include developing a device that can monitor epilepsy outside of the hospital. This would allow it to analyze signals from many electrodes over several weeks or months. Indiveri says, “We want to incorporate low-energy wireless data communications into the design – to link it to a smartphone, for instance.” Johannes Sarnthein is a UniversityHospital Zurich neurophysiologist. He explains: “A portable, implantable chip like this could identify periods when there is a higher/lower rate of seizure incidence, allowing us to deliver personalized medicine.” This partnership includes University Hospital Zurich, the Swiss Epilepsy Clinic, and University Children’s Hospital Zurich.
Comments