IBM trained artificial intelligence to analyze the development of Parkinson's disease

The research arm of Big Blue and the Michael J. Fox Foundation (MJFF) has built a model of artificial

intelligence, which can group typicalParkinson's disease symptom patterns. It can also accurately determine the progression of these symptoms in a patient, regardless of whether they are taking medications to control them.

The report on this discovery was published onpages of The Lancet Digital Health. IBM Research and MJFF have been collaborating since 2018. The goal of the project is to adapt machine learning technologies to help clinical researchers further understand the foundations of Parkinson's disease, especially in the part where the disease progresses differently in different people.

As part of the development of the AI ​​model, researchersused de-identified datasets from the Parkinson's Progression Markers Initiative (PPMI).

“The dataset served as input forapproach to machine learning, allowing the discovery of complex patterns of symptoms and their progression, says a research paper from IBM Research. — Whereas many previous studies have focused on characterizing Parkinson's disease using only baseline information, our method relies on seven years of patient data. Additionally, the model makes limited a priori assumptions about progression pathways compared to previous studies.”

As a result, researchers discovered that the conditionpatient's response may vary depending on a number of factors. Such factors include features of activities of daily living, problems with slow movements, tremors of the limbs, instability in body positions, as well as symptoms not directly related to motor skills: depression, anxiety, cognitive impairment and sleep disorders. In addition, AI has learned to predict the onset of a severe stage of Parkinson's disease. 

As shown by clinical trials, the proposedThe IBM Research model provides fairly high-precision predictions. In the future, other factors will be added to the original data, including genetic information and data from neuroimaging. As the authors of the study note, this will ultimately help to investigate the disease in even more detail.

Read more

New mini-antibodies block coronavirus 1,000 times better than previous ones

Scientists have managed to convert the energy of thermonuclear fusion into electricity

Researchers first created metallic water in terrestrial conditions