BCIwiki (bciwiki.com) — A preprint uploaded to arXiv on 2026-10-16 introduces TriRouter for motor imagery decoding when EEG channels are sparse or corrupted. Motor imagery brain-computer interfaces (BCIs) distinguish tasks by recognizing EEG patterns associated with imagined actions. The preprint has not been peer reviewed. The authors state that it was submitted to IEEE ICASSP 2027; submission does not mean acceptance.
TriRouter is an electrode-space routing framework that allocates contributions from processing paths according to observable channel availability. It combines discriminative-preservation, mask-robustness, and latent-compensation paths with a validation-calibrated router. The authors designed it to preserve discrimination with complete channels while improving representation recovery under severe missingness.
On BNCI2014-001, the authors report that TriRouter outperforms standard public backbones and a dynamic-spatial-filtering corrupted-EEG baseline. Against the study's strongest corrupted-EEG baseline, accuracy improved by 3.68, 4.02, and 6.46 percentage points at nominal missingness settings of 25%, 50%, and 75%, respectively. These figures are gains over the baseline, not absolute classification accuracies.
The authors also report effectiveness under structured regional missingness, where channels are missing in regional groups, and in an additional low-channel dataset evaluation. These tests address classification performance under changing channel availability; they do not establish clinical benefits or performance during practical online use.