Artificial intelligence to speed up the diagnosis of Parkinson’s disease

An AI tool has been created by specialists from the Accelerated Capability Environment (ACE) and the NHS’s AI Skunkworks to expedite the identification of Parkinson’s disease.

The current Parkinson’s disease diagnosis process takes a long time, which restricts the amount of patients that doctors can handle. Patients may be at danger of receiving an incorrect diagnosis, which would cause their illness to worsen before therapy could be started.

Brain alterations can only now be evaluated manually, which takes four to six hours. Additionally, physical brain tissue grading after death is necessary to better comprehend disease reasons and create potential novel treatments, which also takes a lot of time.

In order to better care and therapy for the disorder, researchers have now worked together to build a Parkinson’s disease diagnosis tool driven by AI.

More people are being diagnosed with Parkinson’s disease than any other neurological condition, making it the most common and fastest-growing neurological condition worldwide.

With the number of Parkinson’s patients predicted to double over the next 50 years, this number is anticipated to rise sharply.

The largest risk factor for this neurodegenerative illness is aging, and new diagnostic tools will be crucial in the fight against it.

Using artificial intelligence to speed up the diagnosis of Parkinson’s disease

The largest membership-based charity in the world, Parkinson’s UK, worked with ACE and AI Skunkworks on their study.

The charity’s brain bank at Imperial College London, which has more than 1,300 brains from Parkinson’s patients and healthy donors, was used by the researcher during the 12-week trial.

The nonprofit organization also offered a dataset with 401 digitized images of brain sections that had been immunostained to find alpha-synuclein (a-syn), a pathology marker for the illness. 100 control cases from health donors were also included.

Then, Polygeists from ACE’s Vivace community modified already-existing technology to eliminate certain kinds of brain matter that are not involved in this process. They next used the iDeepColour neural network, which emphasizes regions of the brain impacted by a-syn, to artificially stain slides of brain tissues.

Areas of interest in these photographs are treated such that they seem bright green, making them simple to spot. The green patches could reveal illness density if the photographs were divided up into squares and then broken up.

As a result, Polygeist was able to develop a proof-of-concept classifier with a 92% accuracy rate for diagnosing Parkinson’s disease and no false alarms.

The new AI method drastically sped up diagnosis times by evaluating one brain in only minutes. This frees up neurologists to concentrate on trickier patients.

Upcoming phases of development

The AI tool is almost prepared for usage in practical applications. The group is currently attempting to improve the procedure to distinguish between disease phases and see if any more proteins can be found. The application of this method with live patient brain scans may be viable in the future.