BACKGROUND AND PURPOSE: Early detection of large vessel occlusion (LVO) stroke optimizes endovascular therapy and improves outcomes. Clinical stroke severity scales used for LVO identification have variable accuracy. We investigated a portable LVO-detection device (PLD), using electroencephalography and somatosensory-evoked potentials, to identify LVO stroke. METHODS: We obtained PLD data in suspected patients with stroke enrolled prospectively via a convenience sample in 8 emergency departments within 24 hours of symptom onset. LVO discriminative signals were integrated into a binary classifier. The National Institutes of Health Stroke Scale was documented, and 4 prehospital stroke scales were retrospectively calculated. We compared PLD and scale performance to diagnostic neuroimaging. RESULTS: Of 109 patients, there were 25 LVO (23%), 38 non-LVO ischemic (35%), 14 hemorrhages (13%), and 32 stroke mimics (29%). The PLD had higher sensitivity (80% [95% CI, 74-85]) and similar specificity (80% [95% CI, 77-83]) to all prehospital scales at their predetermined high probability LVO thresholds. The PLD had high discrimination for LVO (C-statistic=0.88). CONCLUSIONS: The PLD identifies LVO with superior accuracy compared with prehospital stroke scales in emergency department suspected stroke. Future studies need to validate the PLD's potential as an LVO triage aid in prehospital undifferentiated stroke populations.