Smart watches could be used to speed up diagnosis of Parkinson’s disease by as much as seven years, using a new AI tool.

People developing Parkinson’s undergo subtle shifts in walking and sleep years before “hallmark” symptoms such as shaking and difficulty balancing appear, but they are not typically picked up until the disease is fairly well advanced.

Using a type of AI known as machine learning, the i reports researchers analysed data collected from more than a hundred thousand smart wrist watches.

They were able to spot those changes, such as reduced acceleration while walking and restless sleep, years earlier than can be done at the moment and while they are still imperceptible to the individual or their doctor.

Dr Cynthia Sandor, a Sêr Cymru Fellow at the Dementia Research Institute at Cardiff University, said “We have shown here that a single week of data captured can predict events up to seven years in the future.”

The watches can’t give a clinical diagnosis, but identifying people at high risk or showing early signs of the disease could aid the process and lead to quicker diagnoses.

There are currently no treatments to slow or halt Parkinson’s, but there are some non-pharmaceutical steps that could help people to slow its progression, although it is unclear how effective they are.

The Parkinson’s Foundation says steps that can be taken include physical exercise, such as walking, cycling, swimming and yoga and eating a whole food, plant based, Mediterranean style diet, including fresh vegetables, fruit and berries, nuts, seeds and fish.

In the near future, smart watches could also help scientists to find a treatment for Parkinson’s by identifying the most suitable people to test drugs on in the crucial early stages of the disease’s development.

This means those people at high risk could be included in clinical trials of new drugs if they wanted to be, while in the future they could access treatments at an earlier stage when they do become available.

Cynthia Sandor said “This is a potentially important, low-cost screening tool for determining people at risk of developing Parkinson’s disease and identifying participants for clinical trials of neuroprotective treatments.

“Machine-learning models achieved significantly better test performance in distinguishing both clinically diagnosed Parkinson’s disease and prodromal [early signs of] Parkinson’s disease up to seven years pre-diagnosis from the general population.”

Currently, there isn’t a specific test to diagnose Parkinson’s disease. Instead, a diagnosis is based on a person’s medical history, a review of their symptoms, and a neurological and physical exam.

Cynthia Sandor said she was surprised and encouraged by her findings.

She said “We did suspect to find some movement or sleep features to be impaired before diagnosis. However, we did not suspect for the impairment to be detectable this early and to achieve better performance than any other risk factors, like genetics or known prodromal symptoms, tested.

“A further surprising aspect we found was the reduction in acceleration to be quite specific to Parkinson’s disease. Other movement disorders did not show this reduction and no other disorder showed it prior to the clinical diagnosis.

“The application area of our tool would lie in the identification of people at risk and in the earliest stages of the disease. Such individuals are crucial for clinical trials of neuroprotective treatments. Instead of testing a new drug on a group of individuals who are diagnosed and thus have already lost about 50 per cent of the dopaminergic neurons, smartwatches would allow us to test a new drug on a group of individuals in the earlier stages of the disease with less neuronal loss.”

She said identifying high risk individuals could make trials smaller and faster.

She used UK Biobank data collected from 103,712 people aged forty to sixty nine who wore a medical grade smart watch for a seven day period in 2013 to 2016 to model whether data from motion tracking devices could be used to identify cases of Parkinson’s disease before clinical diagnosis.

The devices measured average acceleration, meaning speed of movement, continuously over the week long period.

The researchers compared data from a subset of participants who had already been diagnosed with Parkinson’s disease, to another group who received a diagnosis up to seven years after the smart watch data was collected

Cynthia Sandor said “acceleration during light physical activity” was the most significant measure taken by the smart watch. She explained that a slowing of movement, known as bradykinesia, is a key symptom of Parkinson’s and one of the first motor symptoms to develop.

She said she would look to do further research in other groups of people and with other smart watches.

While hopeful that a ‘screening tool’ will emerge, she cautioned that “despite our promising results, there are several further steps until deployment – replicating the results and developing an algorithm that works with consumer-grade products being the biggest ones”.

It was known that Parkinson’s affects movement and sleep, and there are subtle motor symptoms such as a slowing of movement that occur many years before the condition is diagnosed. This study shows a method to detect these in the general population. The researchers also created a machine learning model that could identify these signs and better predict whether someone will develop Parkinson’s.

Scientists who were not involved in the study welcomed the findings although some questioned the benefit to a person of finding out they were at high risk of Parkinson’s in the absence of an effective treatment.

Professor José López Barneo, of the University of Seville, said “The research is very interesting.

“Knowing 10 years earlier that you have a high risk of developing Parkinson’s is very interesting and valuable from a scientific point of view. In addition, the future patient is given the opportunity to prevent/palliate his or her disease.

“However, given that such prevention is not yet possible, it is not clear that this is of any benefit to the future patient. This is an issue with important ethical implications.”

José Luis Lanciego, of the University of Navarra, said “This study has demonstrated that accelerometry measurements obtained using wearable devices are more useful than the assessment of any other potentially prodromal symptom in identifying which people in the normal population are at increased risk of developing Parkinson’s disease in the future, as well as being able to estimate how many years it will take to start suffering from this neurodegenerative process.”

Smart watches are already being used to monitor the progression of Parkinson’s disease and help improve the medication plan but, unlike this new device, these do not pick up signs of the disease far in advance.

The study is published in Nature Medicine and was funded by the UK DRI, the Welsh Government and Cardiff University.

Parkinson’s disease is a brain disorder that causes unintended or uncontrollable movements, such as shaking, stiffness, and difficulty with balance and coordination. Symptoms usually begin gradually and worsen over time. As the disease progresses, people may have difficulty walking and talking.

It affects about a hundred and forty five thousand people in the UK.

Parkinson’s UK’s associate director of research, Claire Bale, said “This could be a major step towards a test that could be used to screen and identify people in the very early stages of developing the condition, years before a formal diagnosis. It could open the door to early intervention with treatments and therapies that can slow, stop or even reverse the damage to brain cells – and potentially prevent Parkinson’s.

“While we don’t yet have treatments that can stop Parkinson’s in its tracks, there are a number of promising candidates on the horizon, with some already being tested in early clinical trials.”