IBM AI predicts Alzheimer’s disease by analyzing descriptions of cookie theft

Researchers at IBM and Pfizer trained the model to identify signs of speech impairment in written descriptions of the image, which is often used to test for cognitive disorders. The transcriptions are from the Framingham Heart Study, which has examined the health of more than 5,000 people and their families since 1948.

The samples were collected before the participants showed symptoms of Alzheimer’s. The researchers then checked the Framingham heart study records to see if a diagnosis had been made later.

They found that the Proof Concept model can predict with 71% accuracy that someone will develop Alzheimer’s disease – seven years before it is clinically diagnosed. IBM said this was significantly better than the 59% accuracy of the standard predictions used on a clinical scale biomedical data of a patient.

[Read: What audience intelligence data tells us about the 2020 US presidential election]

Study co-author Guillermo Cecchi said he wanted to add more geographically, socio-economic and racially diverse training data to the algorithm:

Ultimately, we hope that this research can lead to the development of a simple, straightforward, and accessible metric that doctors can use to assess a person’s risk of Alzheimer’s disease, leading to earlier intervention.

While there is still no effective cure or prevention for the disease – which affects around 44 million people live with worldwide Today – the early detection method could help delay and treat it.

You can read the full study in The Lancet’s EClinicalMedicine.

Published on October 23, 2020 – 13:13 UTC