1817, Jacob British surgeon and polymath Parkinson published a short 66-page pamphlet about a condition little known to the medical community.of Essay on tremor paralysisParkinson described in excruciating detail his observations of six individuals exhibiting varying degrees of the same symptoms: involuntary tremor, difficulty walking, muscle weakness, and slurred speech.

As you might guess, Parkinson was describing the disease that now bears his name. It is a neurodegenerative disease that affects nearly one million adults in the United States. Since then, over the past 200 years, we have discovered the reason behind the disease – the destruction of neurons that produce the neurotransmitter dopamine – and a cure through a variety of innovative means.

However, doctors still struggle to initially diagnose Parkinson’s disease. Especially in the early stages.Scientists are aiming to change that with a new AI-powered device that can help diagnose Parkinson’s from breathing patterns, according to a paper published in the journal last month. natural medicine.

Here is the background— Devices that detect Parkinson’s disease rely on a technology called wireless sensor networks (WSNs), says Dina Katabi, senior study author and professor in MIT’s Department of Electrical Engineering and Computer Science.

WSN was first developed by the US military in the 1950s to track Soviet submarines. Since then, networks of these dedicated sensors have been used to monitor myriad phenomena such as air quality and natural disasters, as well as for military purposes such as gathering information and identifying security threats.

Today, WSN is increasingly expanding into healthcare. This is because it is much more convenient for collecting health data than wearables that people wear for long periods of time to get temperature, blood pressure, or other vital signs. WSNs, on the other hand, deploy groups of sensors that can collect this information from remote locations and relay it for processing.

For example, if you or a loved one lives in a nursing home, sensor smart environments could improve safety and quality of life. For example, in the emergency room, WSNs allow doctors to respond quickly to the moment a patient’s condition deteriorates.

how they did it— Katabi and her team found that WSN can detect Parkinson’s disease by analyzing breathing patterns during sleep.

“During sleep, your breathing changes and follows your sleep cycle,” Katabi explains. “For example, breathing in deep sleep is different than during REM sleep.”

This decision is based on observations made by James Parkinson himself in 1817 and data from recent research. Parkinson’s disease has been found to destroy neurons in the brain stem that control breathing.

Nocturnal breathing changes can reveal early brain changes and disorders, as they can appear years before obvious signs of Parkinson’s disease.

But how can a wireless wall-mounted sensor actually perceive changes that are imperceptible to the human eye?

MIT graduate student Yuzhe Yang and postdoc Yuan Yuan, both members of Katabi’s team, manipulated neural networks that resembled the human brain through connection algorithms that mirrored the behavior of real neurons. . Through his 12,000 nights of sleep data from sleep laboratories across the United States, they taught the network to recognize typical nocturnal breathing and Parkinson’s-related nocturnal breathing.

To detect changes in breathing, neural networks (built into devices like common WiFi routers) use echolocation like bats do.

But instead of sound waves, the device sends radio waves that bounce off the body, says Katabi. These altered radio waves return to the device and provide an image of the sleeping person’s breathing patterns. A neural network then examines the “image” and provides its rating.

what they found — For Parkinson’s patients, the AI ​​focused on breathing during sleep onset and wakefulness. This makes sense because people with Parkinson’s disease tend to be light sleepers and have more interruptions in their sleep. The device can also detect differences in a person’s breathing over time, indicating worsening cases. (This was tested on his large dataset of 7,687 people, including 757 Parkinson’s patients.)

“Neural networks can not only determine if this person has Parkinson’s disease, [but] It also provides an estimate of disease severity,” says Katabi. This is important. Because researchers want to track how quickly the disease progresses and why these differences exist.

Katabi and her team wanted to make sure they were detecting Parkinson’s disease, as other neurodegenerative diseases can cause overlapping symptoms. Based on data from an Alzheimer’s patient, the AI ​​showed a sensitivity close to 81% for her. It had a specificity of 78% when distinguishing between Parkinson’s and Alzheimer’s disease.

Important reasons — Currently, there are no non-invasive methods that can detect the early stages of Parkinson’s disease. Also, in clinical trials, researchers rely on imprecise questionnaires, or a combination of questionnaires and tasks, to determine participant behavior or measure disease progression.

“[These means] It’s subjective or semi-subjective, and it’s not sensitive,” Katabi says. It’s happening.”

There is also the burden of Parkinson’s disease, which is caused by a larger population living longer. Between 1990 and 2015, the number of people with Parkinson’s disease doubled to over 6 million and is expected to double again by 2040 to over 12 million.

The rising incidence of Parkinson’s disease comes at a high price: one data for 2020 estimates the total economic burden of the disease at about $52 billion Nature study. There is no cure for Parkinson’s disease yet, but early detection may reduce healthcare costs.

What’s next – Smart devices already power our homes (and our lives), so this Parkinson’s-detecting AI seems like a very practical gadget to add to the roster.

Katabi also envisions more diagnostic tools coming in the future. For example, AI can recognize the smell of Parkinson’s disease and even distinguish between certain types of medical conditions (a more common sweep at the moment).

It can also be programmed to recognize Alzheimer’s disease, amyotrophic lateral sclerosis and other neurodegenerative diseases that are difficult to detect in the early stages.