Tele-stroke connects an on-call vascular neurologist to an emergency department within minutes of a suspected stroke, because the treatment window is governed by the clock. Eligibility for thrombolysis — the clot-dissolving therapy — is measured against time from symptom onset, so every minute of delay narrows the patient's options. The neurologist works remotely with the bedside team to assess the patient, typically using a cart fitted with a remotely controllable camera good enough to support a structured neurological exam such as the NIH Stroke Scale (NIHSS), alongside review of brain imaging and documentation back into the electronic health record (EHR).
For a product team the defining fact is that latency and reliability are not conveniences here — they are clinical features that move outcomes. The video session has to establish in seconds, not after a reconnect dance, and the architecture must have no single point of failure, because a dropped link during an active stroke assessment is a care failure with real consequences. There must also be a clear, fast escalation path for when video genuinely cannot be established — a phone fallback and a defined protocol, not silence.
The engineering bar is therefore acute-care grade rather than scheduled-visit grade. This shapes concrete decisions: redundant network paths, server infrastructure (TURN/SFU) provisioned for instant connection, remote camera control with low motion-to-photon delay so the neurologist can actually examine gaze and facial symmetry, and rigorous uptime monitoring. The common mistake is reusing a consumer-grade or routine-telehealth stack for an acute-care use case where its occasional dropped call, tolerable for a follow-up visit, becomes a clinical hazard.
