Early detection is important but difficult so far, but now researchers are reporting advances in the ability to diagnose Parkinson's disease. The data can be recorded contactlessly using a small device in the bedroom. In addition to diagnosis, the concept could also be used to study the course of Parkinson's disease and to develop new treatment options, say the scientists.
Muscle tremors, a short gait and a motionless facial expression are among the classic symptoms in the advanced stage: Parkinson's disease, also known as shaking paralysis, is the second most common neurological disorder in humans after Alzheimer's disease. The impairment of movement control is caused by the progressive death of certain nerve cells in the brain in which the messenger substance dopamine is produced. There are millions of people affected worldwide and the numbers are expected to continue to rise.
A problem in the treatment of Parkinson's is the difficulty of early detection. Because it often only becomes clear when motor symptoms such as tremors, stiffness and slowness become apparent. However, these signs often only appear several years after the actual onset of the disease. There are ways to detect Parkinson's disease with the help of brain fluid and imaging techniques. However, these methods are invasive, expensive and require access to specialized medical centers. Simpler methods of Parkinson's diagnosis are therefore in demand.
Parkinson's is associated with specific breathing patterns
The team led by Dina Katabi from the Massachusetts Institute of Technology in Cambridge is dedicated to the development of such concepts. Her approach is based on a well-known aspect of the development of Parkinson's disease: "As early as 1817, the work of Dr. James Parkinson found a link between Parkinson's and breathing. This prompted us to consider the possibility of detecting the disease by breathing,” says Katabi. "Some medical studies have shown that respiratory symptoms manifest years before motor symptoms, which means that respiratory characteristics could be promising for risk assessment before Parkinson's diagnosis," says the scientist.
As part of their study, she and her colleagues have now investigated the extent to which artificial intelligence can detect the signs of Parkinson's disease in people's breathing patterns at night. To do this, they trained a so-called neural network. These are combinations of computer algorithms that mimic how the human brain works. They are capable of learning and, through training, can capture certain patterns in data and assign them to a characteristic. For the study, a neural network was trained to recognize certain signatures in breathing patterns that are linked to Parkinson's disease.
The scientists used data from subjects with known Parkinson's disease and healthy controls as a basis. The breathing patterns in a subgroup were recorded using a chest strap that the subjects wore at night. Another group, however, was contactlessly recorded by a device in the bedroom that looks like a Wi-Fi router. It emits radio signals and can use their reflections to detect the breathing patterns of a sleeping person.
AI recognizes the Parkinson's signature
As the researchers report, the neural network was successfully trained to recognize Parkinson-specific patterns in breathing behavior. This was illustrated by test results on a total of 7687 people, including 757 Parkinson's patients. It was shown that the method can already detect the disease with an accuracy of over 80 percent. It is therefore at least suitable as a hint system that can be linked to further tests. There is also evidence that the concept could be useful in identifying the severity of a person's Parkinson's disease and tracking the progression of the disease over time.
It also became clear that the contactless monitoring system can be used successfully for data collection. This means that there is hardly any effort for patients or nursing staff. "Our study demonstrates the feasibility of an objective, non-invasive assessment of Parkinson's at home and also provides early evidence that this AI model could be useful for risk assessment prior to clinical diagnosis," write Katabi and her colleagues. "In terms of clinical care, the approach can aid in the assessment of Parkinson's patients in traditionally underserved regions, and those who have difficulty leaving home due to limited mobility or cognitive impairment," says Katabi.
The scientists will now focus on optimizing and further developing their concept. As Katabi concludes, she also sees potential for Parkinson's research: "In terms of drug development, the process could enable clinical trials of significantly shorter duration and fewer participants, ultimately accelerating the development of new therapies".
Source: Massachusetts Institute of Technology, professional article: Nature Medicine, doi: 10.1038/s41591-022-01932-x