An Ensemble of Hyperdimensional Classifiers: Hardware-Friendly Short-Latency Seizure Detection With Automatic iEEG Electrode Selection

We propose a new algorithm for detecting epileptic seizures. Our first extracts three features, namely mean amplitude, line length, and local binary patterns that are fed to an ensemble of classifiers using hyperdimensional (HD) computing. These features embedded into prototype vectors representing ictal (during seizures) interictal (between brain states constructed. can be computed at different spatial scales ranging from single electrode up many electrodes. This flexibility allows our identify the electrodes discriminate best between states. assess on SWEC-ETHZ iEEG dataset includes 99 short-time seizures recorded with 36 100 16 drug-resistant epilepsy patients. Using k-fold cross-validation all electrodes, surpasses state-of-the-art algorithms yielding significantly shorter latency (8.81 s vs. 11.57 s) in seizure onset detection, higher specificity (97.31% 94.84%) accuracy (96.85% 95.42%). further reduce 3.74 by allowing slightly percentage false alarms (2% loss). only top 10% ranked algorithm, we still maintain superior latency, sensitivity, compared other finally demonstrate suitability deployment low-cost hardware platforms, thanks its robustness noise/artifacts affecting signal, low computational complexity, small memory-footprint RISC-V microcontroller.