AI reads your medical records and predicts which type of Alzheimer's or Parkinson's you will develop
Alzheimer's and Parkinson's are not single diseases -- they are collections of subtypes, each with its own progression, risk factors and genetic profile. That is bad news for the one-size-fits-all treatment approach, but good news if you can tell them apart early enough. Researchers have now developed an AI system that analyses patient records and identifies five clearly distinct subtypes of both Alzheimer's and Parkinson's, long before the disease reaches its worst.
An AI that spots patterns humans miss
The researchers used a so-called transformer model -- the same technology that underpins large language models such as ChatGPT -- but trained on electronic health records. They fed the system the medical histories of thousands of patients: diagnoses, medications, symptoms, lab results and co-existing conditions. From all of that, the model learned to recognise patterns that precede the onset of Alzheimer's or Parkinson's -- the prodrome, the quiet run-up phase of the disease, before it fully declares itself.
The results were striking. The AI identified five clearly distinguishable subtypes for both Alzheimer's and Parkinson's, each with a unique trail of early signals, accompanying conditions and genetic characteristics. Some subtypes showed early cardiovascular problems, others were marked more by psychiatric symptoms or metabolic disorders. Those early signals were sitting in the records -- sometimes years before an official diagnosis -- but had never previously been recognised as parts of the same story.
Personalised care starts with the right diagnosis
Why does it matter which subtype you have? Because it means your treatment can be different too. If doctors know that your variant of Alzheimer's is strongly tied to inflammatory processes, while someone else's is driven primarily by metabolic dysregulation, they can intervene more precisely -- with the right medication, at the right moment. That is the promise of personalised medicine, and AI is now making that promise more tangible than ever before.
For you, the reader, there is something genuinely hopeful here. Your medical record may already contain clues about your risk of neurodegenerative disease, long before you experience any symptoms. As tools like this are developed further and rolled out in clinical settings, early detection could become routine -- not just for people with a clearly elevated genetic risk, but for everyone. The future of dementia prevention starts with data that already exists. We are only now learning how to read it.