Cohort risk stratification for Alzheimer's trials & memory clinics

Enrich your Alzheimer's cohorts before you pay for confirmatory testing.

Neurobility scores and explains Alzheimer's risk across entire patient cohorts using routine clinical data your sites already collect. Prioritize who goes to amyloid PET or biomarker testing, cut screen failures, and extend recruitment to sites without imaging.

Request a demo → ▶ Watch the trailer
20-minute demo for your team · Sample cohort walkthrough · No commitment
A physician reviewing risk results on a tablet with a patient
Patient #0142HIGH · 0.87
Daily function
Memory score
Age interaction
Education
82%
of patients in ORION's high-risk tier had Alzheimer's, vs. 3% in the low tier
0.96Internal AUC (nested CV)
NACCExternally validated
$0New site equipment
PatentPending
60-second trailer

Watch ORION stratify a cohort in under a minute.

Upload a de-identified cohort. Get risk tiers and per-patient drivers your team can audit.

A clinical team reviewing patient data together
The operational problem

Screen failures, slow enrollment, and narrow site reach drain your budget.

Confirmatory testing is expensive and concentrated at a small number of sites. Without a reliable pre-screen, programs send too many ineligible candidates forward, timelines slip, and recruitment stays limited to imaging-equipped centers.

  • 💸

    Budget lost to screen failures

    Every ineligible candidate sent to amyloid PET or biomarker testing is cost and site time with nothing to show for it.

  • 📍

    Recruitment limited to a few sites

    Community and rural sites often lack imaging, which shrinks your recruitment footprint and makes cohorts less representative.

  • 🔍

    Black-box models don't pass review

    Clinical, scientific, and compliance teams need auditable reasoning. ORION shows the drivers behind every score.

For sponsors & CROs

What are screen failures costing you?

Adjust the sliders to match your study. These are your assumptions, so the math is yours too.

Spent on failed tests today
$1,400,000
Potential savings
$280,000

Illustrative estimate based only on the values you entered. A pilot on your historical data shows what ORION would actually have done.

See it on your data →
How it works

From spreadsheet to risk tiers in four steps.

No IT project. No new tests. A de-identified export is all it takes.

1

Upload

Send a de-identified cohort export from your EHR, registry, or trial database.

2

Map

We match your fields to ORION's inputs. Missing a variable? It works with what you have.

3

Score

Every patient gets a low, moderate, or high risk tier.

4

Explain

See what drove each score, patient by patient, so your team can act with confidence.

Memory score (MMSE)Daily function (FAQ)AgeSexEducationAPOE ε4 (optional)BMICardiovascular history MRIPET scanSpinal tap
Who it's for

Built for organizations that screen, refer, and enroll Alzheimer's patients.

🧪

Trial sponsors & CROs

"We need eligible patients, faster and cheaper."
  • Prioritize who goes to amyloid PET or blood biomarkers
  • Cut screen-failure spend
  • Recruit from sites without imaging
  • Build more diverse, representative cohorts
🧠

Memory clinics & neurologists

"We need to prioritize a growing referral queue."
  • Triage waitlists and referrals by risk tier
  • A consistent, standardized read across providers
  • Auditable per-patient explanations for clinical review
  • Identify trial-eligible patients for research partnerships
The evidence

Rigorously built. Tested on data it had never seen.

Developed with leakage-controlled nested cross-validation on ADNI, then externally validated on the National Alzheimer's Coordinating Center (NACC) dataset.

0.96
Internal AUC
Nested cross-validation, ADNI-1 baseline cohort (n = 819), from a single clinic visit
0.76
External AUC
NACC, never used in training. Consistent across age, sex, and APOE ε4 status
0.90
AUC without MMSE
Reduced model for settings where the MMSE isn't collected
≈
Simple = accurate
A transparent logistic regression matched a tuned random forest
Alzheimer's prevalence by ORION risk tier
ADNI cohort, out-of-fold predictions
High risk
81.6%
Moderate risk
42.5%
Low risk
2.5%
Request a demo

See what ORION finds in your patients.

Start with a 20-minute demo. If it's a fit, we run a paid retrospective pilot on your historical data. No workflow changes, no patient contact.

You get

  • Risk tier for every patient
  • Per-patient explanations
  • Comparison to your known outcomes
  • Written report + review call

We need

  • A de-identified data export
  • About 2 hours of your team's time
  • A signed BAA if needed
  • That's it
Founding pilot partners get preferred pricing.

We're working closely with a small group of trial teams and memory clinics in our first cohort.

Founding partner spots

Request a demo →

Reply within 1 business day

Questions

What people ask before a pilot.

Is our patient data safe?

Pilots run on de-identified data, and we sign a Business Associate Agreement when needed. Your data is used only to produce your results, never to train models for anyone else without your written permission.

Is this an FDA-cleared diagnostic?

No. Neurobility is a research and clinical decision-support tool. It helps teams prioritize who needs further evaluation; it doesn't diagnose and doesn't replace clinical judgment or confirmatory testing.

What if we don't collect every variable?

ORION works with the data you have. A reduced model without the MMSE still reached an AUC of 0.90 in development. We'll tell you up front how your available fields affect accuracy.

Does this require IT integration?

Not for a pilot. A spreadsheet export is enough. Deeper integration comes later, only if the pilot proves its value.

What does a pilot cost?

Pilots are priced by cohort size and scope, and founding partners receive preferred pricing. We'll scope and quote after a 20-minute demo, once we understand your data and goals.

How is this different from other Alzheimer's AI?

Most tools rely on imaging or are black boxes. ORION uses inexpensive routine data, shows the reasoning behind every score, and was validated with methods designed to prevent the inflated results common in this field.

From the founder
"Alzheimer's programs shouldn't be limited to the sites that own a PET scanner. ORION gives any site an auditable risk read from data it already collects."
Amir Kelson · Founder, Orion Science Corporation
🏆 3rd place · NOBIC Gateway Patent pending Externally validated on NACC
▶ Watch Amir present ORION

The right patients are already in your data.

Let's find them. Book a 20-minute demo for your team and see ORION stratify a sample cohort.

Request a demo → Email the founder
Or write directly: amir.kelson@neurobility.org
Request a demo →