Design and validation of distributed EEG signal processing methods for a wearable network of galvanically independent recording units.
Innovation goals ---------------- #### General purpose: This project aims at developing wearable EEG through the use of a completely new measurement setup, namely a **wireless network of galvanically isolated EEG units**. This novel modular approach facilitates a flexible miniaturized design with multi-channel recordings at various positions on the scalp, but also requires new signal processing tools to analyze the EEG signals. To limit the energy consumption of our system we will design distributed algorithm design strategies in which EEG data is optimally compressed, shared, and fused across neighboring EEG units. We will focus on the design of distributed algorithms to remove artifacts and identify epileptiform activity in real time. The new opportunities offered by the system will be used for automated detection of seizures and interictal epileptic discharges from multi-channel EEG in daily life, which is important for diagnosis and follow-up of epilepsy patients. Furthermore, the distributed algorithm design framework will also be generalizable to future modular architectures for other physiological sensing modalities. #### Concrete objectives and criteria: 1. Design of distributed algorithms to remove physiological and extraphysiological artifacts. 2. Design of distributed feature extraction tools targeting epileptiform activity. 3. Design strategies for optimal EEG sensor positioning. 4. 'In-the-field' validation of an epileptiform activity detection algorithm on a cohort of patient equipped with our sensor network. 5. Generalizing/translating the distributed processing framework to other sensing modalities such as bio-acoustics. **Note:** The algorithm design for (1)-(4) will also take the '*special*' EEG signal characteristics due to the targeted modular architecture into account (i.e., floating EEG channels, short inter-electrode distances, etc.) The evaluation of our algorithms will be based on two benchmarks: 1. Upper bound benchmark: This involves a comparison with standard cap-EEG with >20 channels and with centralized processing (we believe we can get close to this upper bound if the EEG-units can be optimally placed, as targeted in objective 3). 2. Lower bound benchmark: To demonstrate the benefit of a multi-unit approach, we will compare with the case where only a single EEG-unit is used, which is the state of the art for miniaturized EEG. **Note:** The evaluation metrics are explained in more detail in the work package descriptions. #### Valorization potential: The outcome of the project will allow Byteflies to offer multi-channel EEG in a unique modular platform to build neuro-related health applications. In addition to this novel EEG sensing architecture, the project will develop distributed algorithms, which can be embedded in Byteflies' Sensor Dot, to remove artifacts and extract relevant biomarkers in EEG, some of them specifically for epilepsy. Byteflies can market these feature-extracting algorithms as part of the library of biomarkers it offers. The project also aims to develop an automatic seizure detector. This can be used as a replacement for the so-called 'seizure diaries' that are used by pharma in clinical trials and by neurologists in patient follow-up, but which are known to be unreliable. The framework of distributed sensor networks will allow Byteflies to rapidly develop new applications in other multi-sensor modalities as well (e.g. in bio-acoustics).