Scientists are developing a neural network capable of detecting Parkinson's disease before it appears

European biologists have developed an artificial intelligence that can detect Parkinson's disease seven years before its first visible symptoms appear by how the concentrations of eight proteins found in patients' blood change.

This was reported on Tuesday, June 18, by the press service of University College in London.

Michael Bartel, a researcher at the University of Göttingen, said that measuring the concentration of these eight protein molecules in patients' blood allows Parkinson's disease to be detected several years before patients feel its first symptoms, and this allows us to start treatment in the early stages of the disease, which in theory will allow us By slowing or even preventing its development.

Experts made this discovery while studying the composition of blood samples obtained from 100 people with Parkinson's disease immediately after identifying the first signs of the development of this disease. The scientists measured the concentrations of various biomolecules found in blood samples and compared the results of these measurements with the amount of these substances present in the plasma of healthy volunteers of the same age and gender, as well as patients suspected of having early forms of the disease.

In order to make such comparisons, the researchers set up a specialized neural network that gradually learned to use small differences in concentrations of blood proteins and other biomolecules to detect Parkinson's disease at different stages of development. The analysis conducted by scientists showed that for diagnosis and early detection of the disease, it is sufficient to use a group of eight proteins found in the blood of patients.

Subsequent observations of the neural network showed that it was able to detect obvious Parkinson's disease with 100% probability based on this set of biomarkers, and also with 80% probability to predict whether the disease would develop in a patient with suspected Parkinson's disease. As the researchers note, their system is able, in some cases, to make such predictions seven years before the first symptoms of the disease appear.

The scientists hypothesized that a neural network could also be trained in a similar way to determine the severity of Parkinson's symptoms, as well as distinguish it from other similar neurodegenerative diseases, such as multiple system atrophy and dementia with Lewy bodies. Biologists concluded that this will improve the quality of diagnosis of these diseases, and will also allow choosing a more correct treatment for their treatment.