SEFMID: a synchronous EEG-fNIRS motor imagery dataset

We introduce and openly release SEFMID, a concurrent hybrid EEG-fNIRS motor imagery dataset collected from 50 healthy participants. The dataset includes resting baseline, left-right hand motor execution (ME), and left-right hand motor imagery (MI) conditions, providing a total of 5,000 valid trials. We also provide single-modality and multimodal classification results, as well as neurovascular coupling analyses, to validate the usability and quality of the dataset. The results show that the multimodal fusion algorithm achieves an improvement of approximately 10% over the best-performing unimodal approach. The public release of SEFMID aims to serve as a valuable resource for advancing multimodal MI-BCI algorithms, investigating neurovascular coupling mechanisms underlying motor imagery, and accelerating clinical translation in neurological rehabilitation.