Alzheimer’s AI can predict the disease with nearly 93 per cent accuracy using more than 800 brain scans, researchers say.

The system identified anatomical changes in the brain linked to the onset of the most common form of dementia, a condition that gradually damages memory and thinking.

The findings build on years of research suggesting AI could help spot early Alzheimer’s risk, predict disease and identify patients whose condition has not yet been diagnosed.

Benjamin Nephew, an assistant research professor at the Worcester Polytechnic Institute in Massachusetts, said: “Early diagnosis of Alzheimer’s disease can be difficult because symptoms can be mistaken for normal ageing.

“We found that machine-learning technologies, however, can analyse large amounts of data from scans to identify subtle changes and accurately predict Alzheimer’s disease and related cognitive states.”

The study used MRI scans, a type of detailed brain imaging, from 344 people aged 69 to 84.

The dataset included 281 scans showing normal mental function, 332 with mild cognitive impairment, an early stage of memory and thinking decline, and 202 with Alzheimer’s.

The scans covered 95 of the brain’s nearly 200 distinct regions and used an AI algorithm to predict patients’ health.

Being able to use AI to help diagnose Alzheimer’s earlier could give patients and doctors crucial time to prepare and potentially slow the progression of the disease.

The analysis showed that one of the top predictive factors was brain volume loss, or shrinkage, in the hippocampus, which helps form memories, the amygdala, which processes fear, and the entorhinal cortex, which helps provide a sense of time.

This pattern held across age and sex, with both men and women aged 69 to 76 showing volume loss in the right part of the hippocampus, suggesting it may be an important area for early diagnosis, the researchers noted.

However, the research also found that the way brain regions shrink differs by sex.

In females, volume loss occurred in the brain’s left middle temporal cortex, which is involved in language and visual perception. In males, it was mainly seen in the right entorhinal cortex

The researchers believe this could be linked to changes in sex hormones, including the loss of oestrogen in women and testosterone in men.

These conclusions could help improve methods of diagnosis and treatment going forward, Nephew said.

More than 7.2m Americans are living with Alzheimer’s, according to the Alzheimer’s Association.

More research is being done to reveal other impacting factors.

Nephew said: “The critical challenge in this research is to build a generalisable machine-learning model that captures the difference between healthy brains and brains from people with mild cognitive impairment or Alzheimer’s disease.”

Radiology AI may improve workflows and patient care, but the technology also brings challenges for radiology departments, research suggests.

A focus issue from the Journal of the American College of Radiology brings together invited research and reviews exploring how AI is being used across different practice types.

Barriers include insufficient infrastructure, strict institutional regulations and a lack of insurance reimbursement, all of which can hamper the integration of AI into routine clinical workflows.

Radiology, the branch of medicine that uses imaging such as X-rays and scans to diagnose and treat disease, is widely seen as one of the fields most likely to be reshaped by AI.

The research includes contributions arguing that workflow improvement is not simply a secondary benefit of AI, but a main determinant of whether a tool succeeds.

Gelareh Sadigh, associate editor for health services research at the Journal of the American College of Radiology, said: “When thoughtfully implemented, AI can complement human expertise and improve efficiency and patient care.

“Successful workflow optimisation requires the integration of AI technology into routine workflows.

“This can be hampered by insufficient infrastructure, strict institutional regulations, and lack of insurance reimbursement.

“Poor integration of AI may degrade workflows, satisfaction, and safety and perpetuate bias in healthcare.”

According to Dr Sadigh, the articles in the focus issue reflect a broader shift in radiology: workflow is not a secondary benefit of AI, but a key factor in whether a tool is successful.

If AI is going to meaningfully help radiology, it must make care delivery better and not more complicated.

Ruth C. Carlos, editor-in-chief of the Journal of the American College of Radiology, said: “This focus issue provides meaningful signposts for AI effectiveness as we navigate a rapidly shifting landscape.”