Using AI to stay one step ahead of dementia

This article is published in association with United Nations.

Vebjørn Andersson has seen first-hand how long a dementia diagnosis can take. As a PhD student and dementia researcher at Oslo University Hospital, he sees many patients with late-stage dementia who may have missed opportunities for earlier support and care.

“It takes too long for the diagnostic procedures to find the right diagnoses for people, and for them and their caretakers to get the help they need,” he said.

Dementia is a condition that affects an estimated 57 million people worldwide. Yet many cases are diagnosed only after significant changes have already occurred in the brain. 

Identifying risk early

As part of the EU-funded AI-Mind project, Andersson and colleagues across Europe have developed new AI tools designed to help identify people at risk of dementia sooner. Earlier and more accurate identification of people at risk could open up opportunities for monitoring, intervention and support. This could potentially extend the amount of time they can live without significant symptoms.

“

By the time cognitive decline has progressed to dementia, opportunities to intervene may already have been lost.Dr Ira Haraldsen, AI-Mind

“By the time cognitive decline has progressed to dementia, opportunities to intervene may already have been lost,” explained Dr Ira Haraldsen, a neurologist at Oslo University Hospital and coordinator of the AI-Mind research. “The point of AI-Mind is to identify the patient early on, before clinical symptoms occur.”

Dementia already affects around 7% of people aged over 60 in EU countries, with the likelihood of developing the condition rising sharply with age. As Europe’s population grows older, finding ways to identify those at risk earlier is becoming increasingly important.

Applying AI

The AI-Mind research team combines medical knowledge with advanced technologies and AI models to improve risk prediction and support diagnostic and healthcare planning. 

The researchers spent more than 50 000 hours collecting and analysing data from 1 000 European citizens using dedicated servers and protocols to keep patients’ information private. This data was then used to develop two complementary diagnostic tools.

One of their main goals was to identify people with mild cognitive impairment (MCI) who are at increased risk of developing dementia later on. Some people with MCI experience difficulties with memory or everyday tasks, but not all will go on to develop dementia. The researchers wanted to identify those most likely to progress.

Studies have shown that changes in the brain’s electrical activity can occur early in the progression of dementia. The first of the diagnostic tools developed, the AI-Mind Connector, uses data on this electrical activity to identify patterns that may be linked to an increased risk of dementia.

Meanwhile, the second tool, the AI-Mind Predictor, uses AI to analyse additional data and further assess the risk of progression from MCI to dementia. It takes into account data from the AI-Mind Connector, alongside cognitive function, biological markers, medical history and background factors such as age, education and where people live.

“

By bringing together all these different aspects, we can identify patterns that help predict a person’s risk of developing dementia over the next two to five years.Vebjørn Andersson, AI-Mind

Reducing risk factors

There is currently no cure for dementia. However, some risk factors such as being socially isolated or drinking a lot of alcohol are modifiable, meaning people may be able to reduce their risk through simple lifestyle changes. Combined with other interventions, this could make it possible to delay or avoid the onset of dementia.

“You can do something about social isolation, you can do something about drinking and smoking habits,” said Haraldsen. As new treatments emerge, she added, tools such as AI-Mind could also potentially help identify people who might benefit from them.

Early identification is about more than treatment, though. It can help people who are experiencing changes in their memory or thinking to understand what may be happening and access appropriate support. According to Andersson, who interacted with many patients throughout the study, many simply wanted answers about their symptoms.

Earlier knowledge can also help people and their families plan for the future, both practically and emotionally.

From research to the clinic

The AI-Mind tools have already been tested in clinical and research settings in Norway. The first phase of the project ended in early 2026, and the researchers are now seeking funding to continue and expand the work to other countries.

Before the tools can be used in medical practice, however, they will need to be approved as medical devices under the EU Medical Device Regulation, which requires a clinical evaluation – something the researchers expect to be completed by 2035.

Testing the algorithms in new environments will help researchers assess how well they work across more diverse populations. The team is also collaborating with industry partners, including Eli Lilly, Roche and Bosch, to further develop the diagnostic approach.

“We have a pretty aggressive strategy to introduce the AI-Mind concept into the medical world,” said Haraldsen. She is particularly hopeful that it could eventually be used in places with fewer medical resources, since the approach does not require highly specialised equipment

“We have a very easy concept which could be delivered on iPads and iPhones, for example,” she said.

And while the tools use AI, Andersson emphasises that they are not intended to replace human judgement or interaction. Instead, they are designed to support conversations and decision-making between healthcare professionals and patients.

“It’s not supposed to make the diagnosis and then leave it up to the healthcare professionals to treat,” he said. “It needs to be a human-based, human-centred decision.”

Research in this article was funded by the EU’s Horizon Programme. The views of the interviewees don’t necessarily reflect those of the European Commission. If you liked this article, please consider sharing it on social media.

This article was originally published in Horizon, the EU Research and Innovation magazine