Photo by Hitesh Choudhary via Unsplash

New Artificial Intelligence Technologies Are Improving Sleep Health and Fighting Disease

By Admin  

According to recent analyses by sleep health professionals, the use of artificial intelligence (AI) technologies is improving sleep medicine by making diagnostic and health monitoring processes more efficient. While these technologies are still in development, they are on track to provide a more streamlined, patient-centered system of analysis for sleep disorders. According to the American Academy of Sleep Medicine (AASM), AI systems allow for more precision and accuracy in sleep tests such as polysomnography, the gold standard of sleep disorder diagnostic tools. At the same time, the speed and efficiency of AI machine learning frees up more time for the specialists who conduct the exams, allowing for more focused and personalized care. Leading CPAP manufacturing companies like Philips Respironics have even developed AI platforms to help the U.S Department of Defense (DoD) in its efforts to fight COVID-19 and reduce the threat of future pandemics. And by using these advanced machine-learning capabilities on existing systems, for example, to access data from smart-CPAP devices or telemonitoring applications, AI can use predictive analytics to track emerging health risks, including nonadherence to treatment. As a rapidly developing technology, AI has the potential to both advance and accelerate sleep health services, giving patients access to more information, better care, and more accurate diagnostic screening.

What is Artificial Intelligence?

Artificial intelligence, often abbreviated as AI, is a term for computer programs that learn through experience and analysis. Referred to as machine learning, this system of improvement over time allows programs to become more accurate and precise by incorporating more data into their processes. This is what is meant by “intelligence,” as the systems are able to “learn” from new information. 

Large AI systems use what is known as “big data” to perform deep analysis of diverse sources of information. This is particularly useful for information that is difficult to categorize or keep updated. For example, a product distribution chain may be large and fluctuate daily or even hourly, requiring real-time data monitoring of numerous information categories. To accomplish this, AI uses sophisticated algorithms to extract specific types of information from the data, and can then use that information to assist in clinical practice. 

In the area of sleep health, for example, one of the most immediate benefits of AI is in the assessment and diagnosis of sleep disorders. Polysomnography, the industry standard for sleep test systems, can be further improved using AI machine learning to fine-tune the test’s accuracy. Ideally, this could further benefit the patient by assessing current health problems as well as the potential for additional comorbid conditions in the future. This is why AI is so useful in sleep health diagnostics, as information assessed during sleep exams is both diverse and highly specialized, consisting of a range of criteria for multiple sleep disorders, often with overlapping symptoms.

AI and Sleep Health

In April of last year, one of the largest sleep health organizations in the country,  the American Academy of Sleep Medicine (AASM) published a position statement on the use of AI technologies in sleep-related medicine. The statement, developed by the AASM’s Artificial Intelligence in Sleep Health Committee, emphasizes both the benefits and limitations of AI in healthcare. 

“While AI applications that score sleep and associated events are expected to improve sleep laboratory efficiency and yield greater clinical insights,” the statement reads, “ the goal of AI integration should be to augment, not replace, expert evaluation of sleep data.” In other words, the advantages of AI are greatest as an improvement of existing systems and protocols, rather than as a replacement for expert opinions. And thus far, sleep health applications are doing well to follow the AASM’s recommendations.

Both of the leading CPAP manufacturing companies, ResMed and Philips Respironics, have developed AI-enhanced data analysis tools to assess patient sleep and health conditions. One example is the Philips Respironics Adherence Profiler, which uses predictive analytics in a cloud-based patient management system. The Adherence Profiler can track patient therapy use and send automated messages when nonadherence begins to be a problem. The messages are sent using telemonitoring software, but they do not replace one-on-one appointments with medical professionals. They simply provide an automated follow-up response, which healthcare providers can use as a prompt for further monitoring and assessment. According to studies that have tested the Philips Adherence Profiler system, the tool is not only successful as an early predictors of therapy adherence, but also allows health professionals to focus therapeutic follow-up efforts on patients who are at risk of therapy adherence problems. And at a time when workloads are increasing in the healthcare industry, tools that help health professionals allocate their time are in high demand. 

The company ResMed is also using AI systems to better manage health data. Currently, ResMed is involved in a joint venture with the healthcare technology company Verily Life Sciences to develop connected-health products focused primarily on sleep health and sleep apnea. Connected health is the term for cloud access to data from a range of devices and applications in the healthcare field. For example, CPAP devices send information to the cloud-based systems that can tell both patients and providers about their use, effectiveness, and any problems that arise. These systems are considered “intelligent” in the sense that they learn from the data and make assessments based on the history of each patient. The more data a system has access to, the more accurate it can become, especially when the data is derived from a diversity of sources (patient adherence, preventative health care, lifestyle, etc.). The goal for many of these systems is to make it easier for providers to diagnose, treat, and manage their patients. And for sleep health in particular, the use of AI helps to establish multi-level strategies to address treatment effectiveness and adherence, an area of concern for sleep health specialists. 

AI and the Pandemic

At the beginning of the COVID-19 pandemic, Philips Respironics began work on an AI-based early warning system that could be used to detect diseases like COVID-19. The new system, called Rapid Analysis of Threat Exposure (RATE), is a part of the DoD’s accelerated response to the current pandemic, involving both the Defense Threat Reduction Agency (DTRA) and the Defense Innovation Unit (DIU) of the DoD. These agencies work to develop and improve technologies used in defense of American lives. Thus far, the program includes SARS-CoV-2 detection among a dozen other pathogens, making it potentially one of the most thorough and effective means of reducing infections nationwide. The RATE system uses large-scale data and machine learning to recognize changes in vital signs and other biomarkers of illness, and with the addition of wearable technologies used as monitors, patients can be tested noninvasively at any time. While the RATE program is only being used by members of the armed services at this time, eventually Philips hopes to advance the project’s reach to the general public, a goal that could help contain the spread of deadly viruses by diagnosing, treating, and tracking patients as early as possible. 

According to information published on the Philips company website, the RATE system works by using AI machine learning and tradespace analysis to assess 165 different biomarkers from a dataset of over 41,000 cases of multiple types of infection, including COVID-19. Philips is still improving the algorithm that measures the physiological changes (biomarkers) and alerts system users of potential dangers. The data, called a RATE score, is collected in the cloud where it’s processed by the algorithm as part of a Software as a Service (SaaS), which allows users to read the scores on a secure website. With the addition of wearable devices such as watches and rings, the system can monitor patients in real time, drastically reducing the need for clinical visits that can further spread the risk of infections. 

The RATE system, like the Adherence Profiler mentioned above, seeks to improve existing infrastructures rather than replace them, making the job of health professionals both easier and more efficient. While traditional approaches to diagnosing infections focus primarily on symptoms that can then be addressed with treatment and/or prevention measures after the fact, the RATE system hones in on physiological responses to infection that can include disease symptoms but can also include more subtle changes in people who are pre-symptomatic or asymptomatic. This level of precision makes prevention measures more proactive, rather than reactive, potentially stopping the spread of infection before it has the chance to do its damage. 

AI and the Future 

Diagnostics and disease prevention are, of course, only two areas of AI development out of a multitude of healthcare applications ranging from simple data processing to large-scale industry-wide analytics systems. But these examples show how current developments in AI data management systems are giving us new insights into patient populations and their health needs. As the AASM’s position statement explains, AI and other new technologies should be used to “facilitate more personalized care,” a goal that many patients find reassuring at a time when fears of impersonalization are on the rise. In the same way, the ultimate goal of machine learning should be our learning, not just as doctors and patients but as a society with serious public health concerns. By using this technology to expand our understanding of sleep and sleep health, we can improve the lives of millions of Americans for generations to come.   

Sources

American Medical Association Journal of Ethics - https://journalofethics.ama-assn.org/article/does-health-information-technology-dehumanize-health-care/2011

Centers for Disease Control and Prevention - https://www.cdc.gov/

Defense.gov - Rapid Analysis of Threat Exposure - https://www.defense.gov/Explore/News/Article/Article/2356086/ai-aids-dod-in-early-detection-of-covid-19/

Journal of Clinical Sleep Medicine - AASM Position Statement - https://jcsm.aasm.org/doi/10.5664/jcsm.8288

Journal of Clinical Sleep Medicine - AI Background Information for Clinicians - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7161463/

Journal of Otolaryngology Head and Neck Surgery - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4992257

MobileHealthNews - https://www.mobihealthnews.com/content/verily-resmed-team-joint-sleep-apnea-venture

Philips Respironics - Care Orchestrator Brochure - https://philipsproductcontent.blob.core.windows.net/assets/20200424/11fe20f1ab454d7a9467aba700e6bc8d.pdf

Sleep and Breathing - https://pubmed.ncbi.nlm.nih.gov/32974833/