Artificial intelligence can be used to detect Alzheimer's six years before the patient is normally diagnosed, says the study.
Doctors used a self-learning computer to detect changes in brain scans too subtly to people to see.
The system was able to identify dementia in 40 patients on average six years before being formally identified.
British AI expert Prof. Noel Sharkey, of Sheffield University, told his findings: "This is exactly the kind of task that deep-seated learning has been cut to find high-quality information models.
"Although the sample sizes and test series were relatively small, the results are so promising that much larger research would be profitable."
Boffins at the University of California trained a computer for scanning more than 1,200 patients.
Scan to measure brain activity by monitoring the intake of radioactive liquid into the blood.
The study has linked the development of Alzheimer's disease to certain brain regions, but can be difficult to detect.
The Alzheimer's algorithm was able to teach itself to identify patterns in brain studies that showed the disease.
The final test included 40 studies of 40 patients who had not previously been investigated.
It turned out to be 100% accurate when detecting Alzheimer's disease many years before the diagnosis of the patient.
Dr Jae Ho Sohn, who worked on the project, said: "We were very pleased with the algorithm's performance.
"It was able to predict every case that advanced Alzheimer's disease."
Alzheimer's early detection could open the door for new ways to slow down or even stop the progression of the disease.
Dr Carol Routledge, Alzheimer's Research UK, said: "Dementia-related illnesses start in the brain for up to 20 years before the symptoms begin to appear and we have an important opportunity to intervene before massive damage.
"This study highlights the potential for learning about language learning in early detection of Alzheimer's disease such as Alzheimer's disease, but the results need to be strengthened to a much larger group of people before we can properly assess the power of this approach."
The study is published in the latest edition of Radiology.
Click More on The Sun.