New Research Uses 3D Facial Imaging to Screen for Sleep Apnea
3D facial imaging technology can be used as a screening tool for sleep apnea, according to research by the American Association of Sleep Medicine (AASM). The study, titled “Predicting sleep apnea from 3-dimensional face photography,” used craniofacial scanner systems (cephalometry) to identify specific physical markers for the screening of sleep apnea. Features such as neck width and retrusion of the lower jaw have been previously associated with sleep apnea, but additional markers such as jaw size, width of the face, and distance between the eyes were also recognized as indicators of risk. Using precise measurements in a prediction algorithm, the researchers were able to assess high and low risk patterns with 89 percent accuracy. And combined with other measurements such as computed topography (CT), these methods can reach an accuracy level of 97 percent or higher. While 3D imaging technology has been used in sleep apnea studies before, the research team hopes these new developments will allow the technology to be adopted for individual, clinical, and mobile health applications on a larger scale, potentially closing the gap between those treated for the disorder and those left undiagnosed.
Sleep apnea is a widespread problem with a high potential for additional comorbid conditions. Yet despite improvements in treatment and more efficient diagnostics, a majority of individuals with the disorder remain underdiagnosed and untreated. Even with continued advancements in PAP therapy technology and numerous features added for patient comfort and convenience, healthcare providers still struggle to get patients started on treatment. The biggest problem, other than treatment adherence, is recognizing the patterns early in order to intervene and initiate the diagnostic process. In recent years, the proliferation of easy-access diagnostic tools like smartphone apps have helped individuals recognize the signs and symptoms of the disorder before the condition becomes severe. While these apps are not highly accurate, they do provide a valuable means of identifying some common signs of sleep apnea. As researchers continue to develop new technologies for diagnostic purposes, it is hoped that improved screening tools will lead to higher treatment rates over time.
Published in the AASM’s Journal of Clinical Sleep Medicine, this new study looks at a rapidly developing technology as a means of predicting sleep apnea, particularly obstructive sleep apnea (OSA), in the general public. The most prevalent form of sleep apnea, OSA is also the most physical in its pathogenesis, resulting from blockage in the airway caused by anatomical traits. Soft tissues such as the tongue and palate, and hard tissues such as the skull and jawbone, can both contribute to OSA blockage. One of the reasons for OSA’s high numbers is the range of morphological features that can potentially contribute to its development.
According to Peter Eastwood, the study’s lead investigator and the Director of the Centre for Sleep Science at the University of Western Australia (UWA), “simple, accurate screening tools are needed to predict those who have OSA.” Making use of ”predetermined landmarks” on the face and neck, Eastwood says that an improved set of predictive markers have been developed with the 3D images, including discrepancy between the upper and lower jaw, elongated teeth, and high upper and lower face heights. This not only makes the technique more effective but also more flexible, as the additional landmarks can be used to more accurately assess the likelihood of obstruction.
Eastwood also worked with Syed Zulqarnain Gilani, a computer science professor at UWA, to help identify features most closely associated with the disorder. Gilani helped to establish the algorithm that separated the data sets (facial measurements) into groups based on the total level of risk. These types of programs are developed using machine learning, where the system becomes more effective over time as the patterns for each classification become more clearly defined. In other words, high risk, low risk, or no risk groups become more easily recognized with each new set of input data. By the time of the study’s publication, the algorithm achieved a 97 percent sensitivity to risk factors, and 76 percent specificity.
3D facial analysis makes use of linear and geodesic measurements. In 3D images, geodesic measurements are the shortest distance between points over a curved surface (such as a skull or other facial feature). These measurements are combined then compared to the set of landmarks to predict the presence and severity of sleep apnea, which is defined according to the apnea hypopnea index (AHI). AHI represents the average number of apnea and hypopnea events per hour. While work is still needed to hone these techniques and develop an efficient method of providing the technology to the general public (as well as promoting its use), it is clear that craniofacial anatomy is an important tool in the effort to recognize and diagnose individuals on a larger scale.
Further Research and Development
The challenge facing more widespread use of 3D imaging tools is how to make the technology more accessible while still maintaining a high level of accuracy. In order to remain efficient in clinical and home settings, image-analysis technology has to become both user-friendly and affordable, as well as adaptable to existing technologies already on the market. Combined with existing technologies like pulse oximetry from mobile apps and wearable tech, 3D scans can potentially become safe, convenient, noninvasive, and inexpensive enough for large-scale adoption. Another option is to integrate 3D scanning into other technologies such as sleep data apps. These apps use optical, acoustic, infrared, ultrasound, or other types of sensors to assess health and sleep on a regular basis. The study by Eastwood and his colleagues helps provide a significant step in this direction, not only for diagnostic purposes, but for long-term health needs and treatment. Because these facial and skeletal landmarks are so accurate as predictors of the disorder, this advancing technology may lead to the first widely used screening tool for sleep apnea risk and assessment.
American Journal of Orthodontics and Dentofacial Orthopedics – https://www.ajodo.org/article/S0889-5406(95)70101-X/fulltext
Dental Press Journal of Orthodonics – https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4686746/
Journal of Clinical Sleep Medicine – Commentary – https://jcsm.aasm.org/doi/10.5664/jcsm.8402
Journal of Clinical Sleep Medicine – Study – https://jcsm.aasm.org/doi/10.5664/jcsm.8246
Sleepapnea.org – https://www.sleepapnea.org/learn/sleep-apnea-information-clinicians/
Springer Link – https://link.springer.com/article/10.1007/s00405-006-0241-5