AI-enabled earwax smell test could help diagnose Parkinson’s: Study
Novel approach might ultimately be screening tool for early detection
A new method that uses artificial intelligence (AI) to analyze the smell of earwax may help detect Parkinson’s disease, a new study from China reports.
This affordable approach could one day be used to screen people for early signs of Parkinson’s, making it easier to catch the disease before symptoms become severe, according to the researchers.
“Further enhancements to the diagnostic model could pave the way for a promising new PD [Parkinson’s disease] diagnostic solution and the clinical use of a bedside PD diagnostic device,” the team wrote.
The study, “An Artificial Intelligence Olfactory-Based Diagnostic Model for Parkinson’s Disease Using Volatile Organic Compounds from Ear Canal Secretions,” was published in the American Chemical Society journal Analytical Chemistry.
Parkinson’s is a neurodegenerative disease, and to date, there are no approved treatments that can stop its progression. If diagnosed early, however, interventions may be able to slow the disease’s progression and enhance a patient’s quality of life.
Early detection, however, has proven challenging. Currently, getting a diagnosis may involve neuroimaging, genetic testing, and neurological examination. This process can be costly and, in the case of physician assessments, subjective.
“To provide accurate and timely treatments for [Parkinson’s] patients, there is a dire need for an objective diagnostic solution that is easy to operate, fast to scan, and low cost,” the researchers wrote.
Research focus on earwax smell as possible diagnostic tool
In the past decade, the olfactory system — a person’s sense of smell — has garnered some attention among other environmental factors as a possible way to identify Parkinson’s. A 2016 study reported that a woman whose husband had Parkinson’s was able to detect the disease using her sense of smell.
Since then, several studies have attempted to replicate these findings with objective diagnostic tools. Most have focused on sebum, an oily substance secreted by the skin that contains volatile organic compounds (VOCs). VOCs contribute to specific odors.
A rapid skin swab test and an AI-based system both demonstrated some promise in diagnosing Parkinson’s based on sebum samples. However, environmental factors can affect VOCs in sebum, making it a potentially unreliable source of information.
In this study, a research team examined an alternate source of VOCs: earwax.
Given that the ear canal is more insulated from the environment, the researchers hypothesized that this sampling method could be more stable.
“Similar to the sebum on the skin surface, ear canal secretions (ECS) contain lipids, proteins, and other compounds, including VOCs, that provide insights into the body’s metabolic status and inflammatory conditions,” the team wrote.
Model showed over 90% accuracy in IDing people with/without Parkinson’s
To test this idea, the researchers obtained earwax samples from 100 people with Parkinson’s and 79 people without the disease. The team then used a technique called gas chromatography to identify VOCs in the samples. This method, combined with specialized sensors, separates components of a sample, including VOCs, based on boiling points and other features.
The researchers identified four VOCs that were significantly different between the individuals with and without Parkinson’s, and could be used as potential biological markers of the disease.
Next, the team trained an AI olfactory (AIO) model with these data and asked it to differentiate between people with and without Parkinson’s. After data processing, the system performed this task with 94.4% accuracy, the data showed.
“The AIO based analytical system underscores its potential for use in bedside medical diagnostic devices, aiding in earlier and more effective treatment for [Parkinson’s] patients,” the researchers wrote.
Two different sensor mechanisms were used to extract results from gas chromatography analysis. One method, mass spectrometry, could more accurately identify specific VOCs, but was slower and more expensive. The researchers suggested that mass spectrometry results could validate results from another method, which used surface acoustic wave sensors. Because of its speed and cost-effectiveness, the latter method might be more useful in a bedside setting, according to the researchers, who also noted that its results supported the high AIO model accuracy rate.
The [AI olfactory (AIO) model]-based analytical system underscores its potential for use in bedside medical diagnostic devices, aiding in earlier and more effective treatment for [Parkinson’s] patients.
Several physiological mechanisms could explain differences in VOCs between people with Parkinson’s and those without, research has shown. The components could arise due to long-term environmental exposures or lifestyle habits that may also raise Parkinson’s risk. Aging and other changes in the body may also contribute to both VOC profiles and the development of Parkinson’s.
In this study, the team could not test possible causes of VOC differences in people with Parkinson’s. The researchers noted that future investigations could compare VOCs in people with Parkinson’s to their family and friends rather than random controls. This might limit the potential effects of environmental exposures and lifestyle, honing in on Parkinson’s-specific changes.
Combined with scales that rate motor and nonmotor symptoms of Parkinson’s, the AIO model could assist in early diagnosis, the researchers also noted. A multifaceted diagnostic approach is most likely to catch subtle changes associated with early Parkinson’s, according to the team. However, the system will require further refinement.
“This method is a small-scale single-center experiment in China,” Hao Dong, one of the study’s author, said in a press release. “The next step is to conduct further research at different stages of the disease, in multiple research centers and among multiple ethnic groups, in order to determine whether this method has greater practical application value.”
Funding for this research came from the National Natural Sciences Foundation of Science, Pioneer and Leading Goose R&D Program of Zhejiang Province, and the Fundamental Research Funds for the Central Universities, per the release.
About the Author
