Researchers at Massachusetts Institute of Technology have developed an artificial intelligence system that can diagnose Parkinson’s disease and its severity by examining a patient’s nightly breathing patterns.
The system, developed by a team at MIT's Jameel Clinic headed by Syrian-born digital health expert Prof Dina Katabi, can detect Parkinson’s earlier than ever before.
Prof Katabi told The National that early detection is vital for managing symptoms of the disorder and reducing patient suffering through the worst of them.
“We mostly diagnose Parkinson’s today through the motor symptoms that it is known for, tremors, stiffness etc,” she said.
“The problem is that the onset of the disease is five to 10 years before these symptoms show, so by the time they manifest it’s too late to do any more than just manage the symptoms.”
Parkinson’s is caused by the death of cells in the substantia nigra, a small region of the midbrain that is responsible for dopamine production, an essential neurochemical for a wide range of functions, including movement and cognitive ability.
Prof Katabi said early-stage diagnostic data for the illness has so far been very limited as it was often expensive and invasive to acquire.
In the absence of more data about its early phase, treatments have been unable to do more than just remove the worst of the motor symptoms for a few hours.
“There is really only one drug that we have for treating PD, however, it is merely symptomatic and does little to stop the actual disease from progressing and patients’ condition just gets worse with time,” said Prof Katabi.
In 1819, English surgeon James Parkinson became the first medical professional to conclusively diagnose the disorder in “An Essay on the Shaking Palsy”.
Prof Katabi found observations in these writings that the breathing of patients changed when they developed Parkinson's, which inspired her team to look further into the matter.
She said there have also been more recent medical studies on PD that proved that respiratory symptoms were evident years before motor symptoms.
To measure patients’ nocturnal breathing patterns, Prof Katabi’s team used a contactless device she had developed nine years previously to monitor the condition of patients with various other illnesses.
The device, called Emerald, emits radio signals around the patient and measures how they reflect off their bodies, allowing it to record important metrics such as heart rate, breathing patterns and whether the patient is sleeping.
The Force meets Superman's X-ray vision
“We discovered that radio waves were kind of like The Force in Star Wars — they are always around unseen. And because we are made up mostly of water, our bodies and their functions are able to affect the way radio waves move around us in measurable ways,” Dr Katabi told The National, as she pointed to a device the size of a Wi-Fi router mounted on her wall.
“Once we understood it that way, it was something akin to Superman’s X-ray vision, we could use the tech to monitor people’s sleeping patterns, their heartbeats. We could basically see people and measure their movements through walls.”
The device feeds all the information it collects into a neural network developed by Ms Katabi’s team, which analyses the data and informs the team whether a patient has Parkinson's and, if so, how advanced it is.
The AI model was trained through the input of data from 11,964 nights, with more than 120,000 hours of nocturnal breathing signals from 757 Parkinson's subjects and 6,914 control subjects.
Ms Katabi said that while the machine knows what to look for in someone’s breathing data to accurately diagnose Parkinson’s, it lacks the ability to inform her team what that key metric is.
“Even if you have the best Parkinson’s expert in the world, you can’t just give them a breathing signal and ask whether this person has Parkinson’s or not,” she said.
“They would not have an answer. The machine can see patterns that are very difficult for us humans to see, because they are sporadic or minute, or because they are a small part of a complex mechanism that we don’t understand yet.”
The fact that the device is contactless and non-invasive will expedite clinical trials for PD treatments significantly.
It allows doctors to monitor patients in a natural state and not while they are hooked up to cumbersome brainwave-measuring machines.
In June, the World Health Organisation said Parkinson’s was the fastest-growing neurological disease, with a prevalence that has doubled in the past 25 years.
Parkinson's affects about 10 million people worldwide and is the second most common neurological disease after Alzheimer’s.