The amount and type of sensory data that can be recorded are continuously increasing due to the ongoing progress in microelectronic technology. Nevertheless, those technologies cannot compete with an actual brain when processing sensory information or interactions with the environment in real-time. “Neuromorphic computing” is a promising approach that has been recently proposed to address these challenges. It connects the functioning of artificial and natural intelligence.
A recent study was conducted using the neuromorphic approach to create a chip that accurately analyzes complex biosignals. It is thought to be an effective method for the diagnostics of epilepsy patients. Epilepsy is the most common severe neurological disease, a central nervous system disorder characterized by abnormal brain activity. It causes periods of unusual behavior, sensations, or loss of awareness. High Frequency Oscillations (HFO) are a biomarker for epileptogenic brain tissue. It can be detected by the analysis of biomedical signals of the brain. An example is an intracranial electroencephalogram (iEEG) analysis, which is used to find problems related to the brain’s electrical activity.
The researchers designed a new algorithm that simulates the natural network of the brain in order to detect HFOs. This system was then integrated into a small piece of hardware with a total area of 99 mm2 that receives signals through electrodes. It also has an advantage in energy efficiency over conventional computers and does not rely on the internet or “cloud computing” solutions. This makes the analysis highly precise and available in real-time conditions.
The proposed system is targeted toward constructing a compact and low-power long-term epilepsy monitoring device that can be used to address a clinically relevant problem. When used as an additional diagnostic tool, the system could improve the outcome of neurosurgical interventions.