EEG-based motor imagery classification with quantum algorithms

Developing efficient algorithms harnessing the power of current quantum processors has sparked the emergence of techniques that combine soft computing with quantum computing. This paper proposes two methods that effectively exploit quantum processors for electroencephalography-based motor imagery classification. The first method is a wrapper feature selection approach based on a quantum genetic algorithm, while the second approach employs a hybrid classical-quantum network comprising a baseline feature extraction network and a variational quantum circuit serving as the classifier. Our evaluation on the BCI Competition IV dataset 2b yielded competitive mean accuracies of 83.82%, 85.56%, and 73.73% for the subject -dependent cross -validation, subject -dependent hold -out validation, and subject -independent leaving one subject out classification approaches, respectively. Notably, our statistical analysis revealed that the hybrid models performed on par with the majority of state-of-the-art architectures, underscoring the practical viability of quantum -hybrid methodologies in real -world problem -solving scenarios.