Early insight into Alzheimer's, powered by AI
AURORA combines clinical test scores with MRI brain imaging to identify patterns associated with Alzheimer's disease — grounded in real research data, built for clarity.
A research tool — not a medical device. It cannot diagnose Alzheimer's disease or any other condition.
A multimodal model trained to detect Alzheimer's earlier.
Two data streams, one prediction
AURORA fuses clinical assessment scores with MRI brain imaging using a late-fusion model — capturing what neither stream alone can see.
Lightweight, hospital-ready
Built on EfficientNetB1, the imaging pathway can run on modest hardware — early detection shouldn't require a specialised research lab.
Real clinical data
Trained on the ADNI research dataset — real patient MRIs and clinical records — not synthetic samples or an abstract benchmark.
Cognitive test scores. Patient history.
AURORA reads the full clinical picture from structured data — MMSE, CDR, MoCA, ADAS-Cog, Trail Making, and more — to understand where a patient stands.
MRI carries signals invisible to the eye.
AURORA's EfficientNetB1 imaging pathway extracts slices per scan and learns to recognise the structural patterns that precede Alzheimer's onset.
Neither stream alone tells the full story.
AURORA's late-fusion model combines both pathways — weighting clinical and imaging predictions — into one classification with probabilities, more complete than either source alone.
Bring your test results — or start from scratch.
AURORA research analysis
Enter clinical scores a doctor has already measured — MMSE, MoCA, ADAS, Trail Making and the rest — and optionally add a brain MRI. This runs the full multimodal model: classification, probabilities, contributing factors, and a plain-language explanation.
- Clinical scores + optional MRI
- Contributing factors and out-of-range flags
- AI assistant to talk the result through
The 10-minute memory check
If you don't have clinical results, this asks a set of questions and simple thinking tasks at home. It scores itself in plain language, and can pass what it measures to AURORA for a second opinion.
- Nothing needed but yourself
- Runs entirely in your browser
- Feeds AURORA, with every value labelled
New to this? Read about AURORA — how it was built, what it was trained on, and where it falls short.
Built for patients, families, and researchers who need answers.
Science-backed. Research-grounded. Built to inform.
Early detection focus
AURORA is built around the window before symptoms become severe, when a conversation with a doctor still has the most room to help.
Multimodal by design
Fusing clinical data with MRI imaging captures dimensions a single-modality model misses entirely — not just a bigger version of the same signal.
Evidence-based
Every prediction traces back to real ADNI patient data — not synthetic samples or an abstract benchmark.
Accessible architecture
EfficientNetB1 keeps the imaging pathway lightweight enough to run without specialised, high-end GPU infrastructure.
Research transparency
The methodology, dataset choices, and documented limitations are public — so what you're trusting is something you can actually evaluate.
An assistant, not just a score
A built-in assistant answers questions about a result, memory, and Alzheimer's disease in plain language — grounded, not generic.
Three steps from data to something you can act on.
What this can and can't do
What it can do
- Give you a rough sense of how your memory and thinking are doing today
- Point out answers that are commonly worth mentioning to a doctor
- Help you put words to something you've noticed but couldn't describe
- Answer questions about memory, ageing, and Alzheimer's disease
What it can't do
- Diagnose Alzheimer's disease, dementia, or any other condition
- Rule anything out — a reassuring result does not mean nothing is wrong
- Replace a proper assessment by a doctor or memory clinic
- Account for tiredness, stress, medication, hearing, or eyesight, all of which change how anyone performs
Memory changes for many reasons that have nothing to do with Alzheimer's — poor sleep, thyroid problems, depression, vitamin deficiency, and some common medications all affect memory, and many of them are treatable. That is one of several reasons a doctor is the right next step, not a website.
Questions people ask
Is this a medical test?
No. It is not a medical test and it cannot diagnose anything. It is a set of questions and simple thinking tasks that may help you decide whether to talk to a doctor. Only a doctor can diagnose a memory condition.
What if my result looks bad?
Try not to read too much into it. Doing poorly on a short online check is not a diagnosis, and many ordinary things — tiredness, stress, a noisy room, poor sleep, some medications — affect how anyone performs. The right next step is the same either way: book time with a doctor and tell them what you've noticed.
What if my result looks fine?
A good result does not rule anything out. This check is short and rough, and it can easily miss real problems. If you or the people around you have noticed changes in your memory, see a doctor regardless of what this page says.
Do you save my answers?
No. Your answers are used to work out your result and are not stored on our servers. There is no sign-up and we do not ask for your name, address, or date of birth. If you use the question-and-answer assistant, what you type is sent to an outside AI provider to generate a reply, so please don't type personal health details there.
How long does it take?
About 10 minutes. You can stop at any time, and nothing is submitted until you finish.
What is AURORA?
AURORA (Alzheimer's Unified Representation for Omni-modal Risk Analysis) is the research model behind the site's Alzheimer's analysis. It was trained on the ADNI research dataset and can combine clinical test scores with MRI brain scans. It is used for the detailed analysis option, which is aimed at people who already have results from a doctor — not for the at-home memory check.
Who built this?
It was built as a science fair project by a Grade 10 student, and it has not been tested in a hospital or approved by any medical regulator. That is exactly why it should not be treated as a medical opinion. It is offered free, in the hope that it nudges some people toward a conversation with their doctor sooner.