The evidence is settled. The delivery is broken. Across a fragmented pathway, the right decision is lost at every handoff. HORUS is the intelligence layer that closes those gaps, so every patient gets the right treatment, in real time.
A University of Pittsburgh–licensed company
HORUS NeuroTechnologies
HORUS NeuroTechnologies is a University of Pittsburgh–licensed company building the intelligence layer for acute ischemic stroke care. Our systems read the clinical picture in real time and return guideline-graded, personalized decisions at the moment they matter most.
Founded by a vascular and interventional neurology team, and researchers who know code strokes to the core, HORUS turns the evidence the field already has into decisions that reach every patient, at every center, every time.
The problem
Stroke volume keeps climbing while the window to act stays measured in minutes.
Hub-and-spoke transfers split the decision across sites, teams, and shift changes.
The guideline runs to dozens of branching rules. No one recalls all of them under a running clock.
A missed indication is an unforgiven mistake, clinically and for the people who carry it.
Today the pathway breaks at every handoff. Each red mark is a point where the right decision can slip.
The human cost
The clinicians carrying these decisions are paying for the gaps. The evidence is quiet, peer-reviewed, and hard to unsee.
A day in acute stroke
13:47 on a Wednesday. It starts calm. Then a transfer comes in and the questions arrive faster than anyone can answer them. Here is that hour, from the quiet before to the outcome the next day.
The ecosystem
Two products are live and validated in clinical settings today, licensed from the University of Pittsburgh. Eight more are in development across the complete stroke pathway, from the ambulance to outpatient prevention.
Live
The guideline engine for acute stroke.
What to do. Why. In under five seconds. LUCID walks a code-stroke case through eight sequential care stages, carrying state forward at each step, to an evidence-graded recommendation, with the Class and level of evidence attached and the scribe note written for you.
Perfect across the board (95% CI, 98.3–100%). Three frontier language models trailed far behind — 71–78% overall concordance, all P<0.001 vs. LUCID. See the full benchmark →
LUCID runs alongside the code-stroke call, whether at the spoke, in the ambulance bay, or on the hub side of a transfer. The clinician confirms what is already known; LUCID checks every contraindication against the current guideline and returns a graded recommendation plus a ready-to-paste scribe note, so the chart is written by the time the decision is made.
Audited against the 2026 AHA/ASA acute ischemic stroke guideline across 225 decision points, with a validation study underway at two UPMC centers and a real-world retrospective review in progress.
Live
Finds the patients trials miss. In real time.
The first stroke-specific, real-time trial screening system. Paste a clinical note; TrialLens reads it the way a screener would and returns ranked trial recommendations in under 30 seconds, inside the acute treatment window.
100% screening sensitivity after reinforcement-learning tuning, on consecutive presentations.
TrialLens reads the same note the team is already writing and surfaces every open protocol the patient could enter, ranked by fit, before the enrollment window closes. No separate data entry, no screening checklist; the match arrives while the patient is still in front of you.
Evaluated on 143 consecutive presentations across two UPMC centers over 18 weeks, reaching effectively 100% screening sensitivity after reinforcement-learning tuning.
A product walkthrough is available on request.
The whole picture
Decision intelligence, trial matching, and operational automation on a single platform. One integration, one login, one partner across every stage of the pathway.
Guideline-graded recommendations at every care node, with the Class and level of evidence attached and the scribe note written. Powered by LUCID.
Real-time, stroke-specific screening that surfaces every open protocol a patient could enter, ranked by fit, before the enrollment window closes. Powered by TrialLens.
Quality reporting, billing integrity, records, and follow-up handled downstream, so the work the guideline creates does not fall on the team.
Detection tools solve one slice. HORUS covers the whole pathway, end to end.
The evidence
When the clock is running and the room is loud, LUCID does not offer an opinion. It returns the answer, the reason, and the grade, fast enough to matter.
A single graded recommendation for the case in front of you, not a list of possibilities to weigh under pressure.
Every recommendation arrives with the guideline Class and level of evidence attached, so you can defend it, and chart it.
Every contraindication, checked against 225 decision points of the current guideline, every single time. No fatigue, no shift change, no missed indication.
The difference between a fluent guess and a recommendation you can sign your name to.
Head-to-head on the acute-stroke decision tasks that matter. LUCID reached 100% across every category; the frontier models did not.
Per-node concordance was 100% across all eight care nodes (Prehospital 27/27, ED Eval 30/30, IVT 36/36, EVT 36/36, Posterior/MeVO 24/24, Post-Reperfusion 24/24, Antithrombotics 24/24, Inpatient 24/24). All differences P<0.001 vs. LUCID. Presented at ESOC 2026.
The category
Detection tools find the clot. Prediction tools score the risk. General models talk. None of them make a guideline-graded treatment decision and hand back a defensible recommendation. HORUS does.
| Capability | AI detection | LLM tools | Prediction | HORUS |
|---|---|---|---|---|
| Guideline-based treatment decisions | ✕ | ✕ | ✕ | ✓ |
| Evidence-graded output (Class / LOE) | ✕ | ✕ | ✕ | ✓ |
| Real-time clinical-note reading | ✕ | ✕ | ✕ | ✓ |
| Clinical trial matching | ✕ | ✕ | ✕ | ✓ |
| Validated in clinical settings | ✓ | ✕ | ✕ | ✓ |
| Full stroke-pathway coverage | ✕ | ✕ | ✕ | ✓ |
| FDA CDS-exempt, deploy now | ✕ | ✕ | ✕ | ✓ |
News & milestones
Co-founders Raul G. Nogueira and Mo Doheim presented HORUS at the European Stroke Organisation Conference Innovation Session in Maastricht, making the case for an intelligence layer that delivers guideline-graded decisions in real time.
HORUS executed an exclusive, worldwide software license with the University of Pittsburgh covering LUCID and TrialLens, effective May 26, 2026, moving both validated systems from the research bench toward clinical deployment.
Why it matters
Stroke care that misses the window is increasingly ending up in court. "Time is brain" is now a phrase juries hear, not just a line in the guidelines.
A 41-year-old arrived by ambulance; with no board-certified radiologist on hand, inexperienced residents misread the scan. The clot sat for more than three hours. At trial, counsel repeated two words: time is brain. The largest medical-malpractice verdict in the county's history.
Gair, Gair, Conason et al. · read the case →A 52-year-old reached the ED about 30 minutes after onset, well within the 4.5-hour tPA window. The team relied on a wrong "last known normal" and a misread CT perfusion scan, ruling out both tPA and thrombectomy. He was left paralyzed. The jury found the failures entirely preventable.
Finz & Finz, P.C. · read the case →The team
Co-Founder & CEO
Clinical Research Assistant Professor of Neurology, Director of the NeuroPrecision Lab, UPMC Stroke Institute.
Senior Co-Founder & CMO
Endowed Professor of Neurology and Neurosurgery, Director of the UPMC Stroke Institute, Chief of Cerebrovascular Medicine.
In collaboration with
Get in touch
For clinical demonstrations, validation data, or partnership, reach the founding team directly.
horusneurotechnologies@gmail.com