Parkinson’s disease: how neural networks help doctors establish the right diagnosis and make the disease develop more slowly

Today James Parkinson was born. He is the English doctor who described the disease which was later named after him. On the WHO initiative, 11 April is the World Parkinson’s Day. Neurodegenerative diseases, especially with the post COVID-19 syndrome in the picture, require early diagnosis and prevention, and this has become even more obvious. In our review we will show how neural networks help doctors establish the right diagnosis and make the disease develop more slowly.

Early diagnosis based on skin

Health professionals and mathematicians from the USA developed an AI system which determines Parkinson’s disease at early stages based on pictures of stem cell from a patient’s skin. The algorithm was taught based on images of skin stem cells from 91 person with Parkinson’s disease at its different stages and from healthy USA residents of the same age and sex. Subsequent inspections of the neural network functioning showed that AI accuracy was 80%.  Researchers also hope that their development will accelerate search for “targets” for drugs capable of suppressing Parkinson’s disease in most people who have it.

Estimating probability of disease development and drug effectiveness based on eye movement

 EyeMove mobile app calculates the eye movement trajectory, analyzes the response time, time of movement, etc. Green dots appear on a smartphone screen. A person’s eyes should follow them. The eye movements are recorded, so the position of the head, eyes and pupils is determined. The results are compared with normal values and with previous test results. Therefore, a doctor can remotely control the disease development and, if necessary, invite a patient to the office to correct or change treatment.

A video with eyes to create new drugs

The company Neuralight from Israel created a technology for accelerating the development of drugs for Parkinson’s disease, Alzheimer’s disease and multiple sclerosis treatment. A doctor records a five-minute video of a patient’s eyes. Neuralight visualization tools clear and improve the video, and then AI interprets eye movement data.

Neuralight analyzes about 100 parameters, including blinking frequency, speed with which a patient’s eyes stop at a particular object and pupil dilation speed (the latter has a strong correlation with Parkinson’s disease). The company hopes that FDA will give its initial approval by the end of 2022, and the first commercial approvals will be signed in 2023.

Five minutes for interpreting brain MRI

SpinTech, a startup from Michigan, created STAGE platform for quick and accurate diagnosis of neurological disorders.  It can be installed on any MRI system. STAGE needs about five minutes to provide 16 separate parameters, including neuromelanin content in the brain, which helps radiologists diagnose Parkinson’s disease.