AI-powered microscope predicts brain disease formation - study

EPFL scientists have developed an AI-based system that detects, analyzes, and predicts in real-time the formation of key factors in the development of Alzheimer's, Parkinson's, and Huntington's diseases. The innovative approach allows researchers to pinpoint the onset of pre-disease changes with 91% accuracy, writes UNN with reference to Phys.

Details

The accumulation of misfolded proteins in the brain is a key factor in the progression of neurodegenerative diseases such as Huntington's disease, Alzheimer's disease, and Parkinson's disease. But to the human eye, proteins that are destined to form harmful aggregates are indistinguishable from normal proteins.

The formation of such aggregates also tends to occur randomly and relatively quickly – within minutes. The ability to identify and characterize protein aggregates is crucial for understanding and combating neurodegenerative diseases.

Now, using deep learning, EPFL researchers have developed a "self-guided" imaging system that uses multiple microscopy techniques to track and analyze protein aggregation in real-time – and even predict it before it begins

In addition, the approach minimizes the use of fluorescent labels, which can alter the biophysical properties of cell samples and hinder accurate analysis.

This is the first time we have been able to accurately predict the formation of these protein aggregates. Since their biomechanical properties are linked to diseases and impaired cellular function, understanding how they develop throughout the aggregation process will lead to the fundamental understanding needed to develop solutions

In the new study, scientists improved this system, and it now works in real-time: as soon as the algorithm detects a mature protein aggregate, it automatically activates a special Brillouin microscope. This microscope analyzes the properties of the aggregates, such as their elasticity. Previously, the microscope worked too slowly for such tasks, but now, thanks to AI control, it is activated only when truly needed. This allows for quick and accurate observation of protein changes.

According to Ibrahim, this is the first scientific work that shows how self-guided systems can be applied in microscopy without the use of fluorescent labels – and make this approach more accessible to biologists.

This opens up new possibilities for studying toxic protein aggregates associated with brain diseases such as Alzheimer's or Parkinson's. As Professor Lashuel notes, such intelligent microscopy will help create more accurate drugs and bring personalized medicine closer to patients.

Genetic test can predict obesity in childhood - study7/21/25, 3:00 PM • 3583 views