The SLEEP 2020 virtual conference was anything but a snooze festNote: To access the full presentations mentioned below, you must have registered for SLEEP 2020. On-demand presentations will expire in August 2021. How riveting was the SLEEP 2020 conference? It’s quite ironic that our team missed a lot of sleep as the whirlwind of ideas kept us up into the wee hours over the weekend. While many on-demand sessions still warrant our attention in the coming weeks, every live and on-demand session we had the pleasure of attending more than lived up to the pre-show hype. The key theme from the opening SLEEP 2020 discussions was this: as an industry, we’ve moved past the question: “is AI part of the future of sleep?” We’re in a place today where with the right data, the right methods, and multidisciplinary collaboration, AI can solve problems that have challenged both clinicians and computers for decades, elucidate insight into sleep’s basic scientific problems we previously thought intractable, and give rise to new clinical sleep applications, opportunities, and challenges we have still yet to identify. With that said, let’s jump into a few quick takeaways from live and on-demand sessions from the SLEEP 2020 virtual conference.
The Relationship Between Sleep and Alzheimer’sThe opening plenary session took on an incredibly pressing topic, analyzing the intersection of Alzheimer’s and Sleep. Session speaker, Dr. David M. Holtzman, MD, shared findings on how Tau and amyloid beta aggregation leads to Alzheimer’s disease. He mentioned that these aggregations can start showing up in data collected from the brain up to 20 years prior to initial symptoms. Then he discussed how sleep deprivation can ultimately lead towards increased risk for Alzheimer’s based on the relationship. A commentary from one of our engineers: “From the conversation, it seems like the bi-directional relationship could be viewed as a positive feedback loop: sleep deprivation leads to increased tau and amyloid beta aggregation which in turn leads to increased sleep deprivation cycling over and over endlessly.” The question that remains is simple: what types of treatments and early diagnostic procedures can be developed in order to break this loop? Catch the full session on demand by searching: “Sleep and Alzheimer’s Disease: Bi-Directional Relationship With Amyloid-ß and Tau.”
PSGs and Beyond: Where will Polysomnography be in 2030?The session featured a pair of interesting speakers, kicked off with notes from Jennifer Newitt, MD and closed by Charles Atwood, MD. The quote we pulled out to share is from Dr. Atwood.
“Human scoring is not inaccurate. It can actually be very accurate. It simply is not consistent, especially between scorers.” – Dr. Charles Atwood, MD.His overall discussion paints accuracy in a different light. To close the session, Dr. Atwood proposed the following hypotheses regarding PSGs in 2030:
- Sleep labs will still exist!
- Clinicians will use a blended model of technology leveraging newer methods to gain deeper insights.
- Health systems will use more sophisticated data to better understand sleep disorders.
The Future of Scoring and the PSGAnother interesting look at AI’s impact on the future came from Dr. Richard Berry, of the University of Florida. Dr. Berry expressed this sentiment on the expansion of AI scoring when it comes to PSGs: “Let me do what I’m good at and computers do what they’re good at – together, we’re greater.” In short, let the doctors and technologists get into details with patients and treatments rather than spending all their time on scoring. Equally interesting was his commentary on creating a “Sleep Study Cohort for Validation,” which would allow rising sleep researchers and companies to access accurately scored sleep study data to accelerate the impact of AI and computers in PSG scoring. To hear more from Dr. Berry, including the how and why of that concept, watch the session on demand by searching: “The Future of Scoring/PSG.”
AI and Sleep Medicine: The Next FrontierThis session is near to our hearts and features four of the most innovative minds in sleep medicine: Dr. Allan Pack, Dr. Emmanuel Mignot, Dr. Nathaniel Watson, and our own CEO and Co-Founder, Chris Fernandez. Fernandez kicked things off with an overview of AI and how it’s intersecting with the world of sleep. He then posits that artificial intelligence in general is the “theoretical tools and computer programs that address complex reasoning, problem solving, knowledge representation, planning, learning, natural language processing, machine perception, robotics, affective computing and general intelligence.” Fernandez outlines the history of AI and the preceding technological breakthroughs that made its powerful capabilities possible. Comparing sleep and AI to the theoretical-experimental collaboration that ultimately furnished the discovery of the Higgs-Boson, and later achievement of the Nobel Prize. Fernandez outlines the foundational methods, developments, and basic principles of machine learning and deep learning approaches that have shown recent promise in clinical and research applications for the study of sleep, key challenges and considerations for the research, development, and validation of AI applications in sleep, useful ontologies and frameworks for critically thinking about the key areas in which artificial intelligence will be used to support patients and clinicians, and common-ground tools, nomenclatures, and multi-interdisciplinary research collaboration best practices that are needed to truly make the impact patients deserve. The take-home message was clear: the present Clinician-AI collaboration opportunity to sleep is akin to the partnership that resulted in the discovery of the Higgs Boson. But instead of billions of dollars, tens of thousands of engineers, and decades of development, thanks to the amazing foresight of our field’s founder, Dr. Willam C. Dement, this work can be done with affordable supercomputers in the cloud using sleep data we already have, can readily collect, and that we are inspired to create today.
Fernandez concluded, “Together, through collaboration, we can reveal a deeper understanding of the true nature of sleep.”The following three presentations by collaborators at University of Washington, UPenn, and Stanford jump into specific examples of machine learning’s current and future impacts. Allan Pack, MD, followed up with a riveting commentary on the importance of phenotyping sleep disorders using clustering – a collection of unbiased data-relationship discovery approaches in unsupervised machine learning. Per Pack, “In today’s age, unbiased discovery approaches are pivotal. We need to throw out everything we know and let the data speak for itself.” In layman’s terms, Dr. Pack’s team believes their current data-powered AI can better identify which of the common groups of sleep disorders people typically fall into, for example: Insomnia, Asymptomatic, and Excessively Sleepy groups. This concept whereby data is grouped in a way that subjects are consistently similar within groups, and consistently different between groups, is known as clustering. By accurately predicting these groupings earlier in the process, we may be able to provide better and earlier knowledge on what types of treatment would be most effective using nuanced phenotypes of OSA that go beyond the current diagnostic paradigms.
“Clustering, when thought about in an unbiased way, is a very powerful tool,” noted Dr. Pack.Then, Dr. Emmanuel Mignot presented information highlighting the ability of AI to analyze high density data like EEG signals recorded during a PSG study in a superior manner compared to manual visual inspections done by humans. Furthermore, he noted that AI can reveal the areas in PSG studies where human consensus is least likely by producing a range of probabilities associated with different types of events and sleep stages. To end his discussion, Dr. Mignot presented recent research that revealed the possibility of using machine learning to diagnose narcolepsy from PSG studies. To close the session, Dr. Nathaniel Watson discussed uncovering new species of events we can identify and even AI score in sleep studies, to new diagnostic indices like Brain Age that give us a richer picture of a patient’s health and conditions, to the host applications that will spawn from insights into the EEG derived “Encephalome,” the brain’s waveform analogy to the Genome in molecular neurobiology and pharmacogenomics. And to close out a note on this excellent session, we’ll include this commentary from Dr. Watson (slight paraphrase for brevity): “I think AI is presenting a question like the one Neo faces in The Matrix when deciding between the red pill, and the ‘real’ world, or the blue pill, and life in ignorant bliss. The reality is that 70 million Americans suffer from a sleep disease, and 60% of them have a sleep disorder. And the reality is that there’s one board certified sleep doctor for every 45,000 citizens and we’re not putting enough into the field to replace those that are retiring. And although telemedicine does help with geographic issues, it doesn’t necessarily increase access because providers still need to interact with patients on a 1:1 scale, so it doesn’t significantly save clinician time. And all this is while 80-85% of sleep disordered breathing is going undiagnosed and untreated. So, I’m going to suggest we swallow the red pill and embrace what artificial intelligence has to offer the field of sleep medicine. This will help us get more people with sleep diseases diagnosed and treated and feeling better.”
These four sessions are broken into individual units, catch them in the original order by searching for “Artificial Intelligence and Sleep: The Next Frontier.” And if you’d like to learn more about the final two discussions, we did cover similar presentations in more detail in these two pieces: for Dr. Mignot and for Dr. Watson.
Sleeping with Our Emotions: Novel Insights Regarding Sleep-Associated Emotion Recognition, Regulation, and MemoryThis session featuring Tony Cunningham, PhD, Lauri Kurdziel, PhD, and Shengzi Zeng, PhD student launches into emotional memory consolidation when sleep-deprived. Dr. Cunningham analyzed the impacts of traumatic events when fully rested versus when sleep deprived, citing a study by Vargas et. al. 2019. As you could likely guess, sleep deprivation leads to significant issues with memory loss, articulated here by Dr. Cunningham: “The study found that encoding and retrieving information in a sleep deprived state significantly impairs all aspects of memory.” For our nappers out there, Dr. Kurdziel shared an important note about the value of those 20-minute sleep afternoon snoozes: “The study also notes that a brief period of recovery sleep, (aka a nap), not only preserves memory for neural information but ENHANCES their ability to discriminate between emotional content.” The session really gets you wondering how sleep treatments may be able to affect not only cognitive functions, but also social-emotional functionality as well. Dive into the slides: “The Effect of a Nap on Emotional Reactivity in Individuals with a Chronic Mild Traumatic Brain Injury” by Dr. Kurdziel and “One Night of Sleep Continuity Disruption Does Not Impact Emotional Perception, Attention, or Memory,” by Dr. Cunningham.
Careers in Sleep Technology for the Sleep Technologist/Respiratory TherapistThe career-focused session highlights the likely expansion of sleep care in the next decade, specifically looking into the CCSH credential and how it can impact the financial costs associated with OSA. According to the session, the number of diagnosed people in the US (according to 2015 data) is around 5.9M, with another 23.5M who are undiagnosed. Right now, the costs associated with the diagnosed cases are just $12.4B compared to $149.6B in the undiagnosed community. Because of the other health implications that arise as comorbidities with OSA, the average cost per person is more than 3x higher for undiagnosed individuals ($6,366/person) vs. diagnosed individuals ($2,105/person). The sessions presented by Amber Allen, Matthew Anastasi, Thomas Arrington, Grace Denault, Sarah Zallek, and Kelly Gladden, explored Certification in Clinical Sleep Health (CCSH) credential in depth, outlining five key learning objectives for career advancement in the sleep industry in 2020:
- Learn the blueprint of the CCSH exam and all materials that should be considered for preparing for the exam.
- Research and discover career development pathways and what steps you need to take to get there.
- Improve techniques for communicating with patients and educating family members.
- Continually practice reviewing methods to improve adherence to therapy and testing treatments.
- Gain knowledge regarding the sleep navigator and its impact on screening inpatients for potential OSA.
While I was Sleeping: 10 years of Sleep Technologist Change and the Transformation from ‘Trade’ to ‘Profession’In the final session, Matthew Anastasi discussed new and trending careers in sleep. New and evolving positions include: sleep navigator, sleep health educator, pulmonary clinic RPSGT, pre-auth specialist, sleep program ancillary staff, sleep coach (behavioral sleep medicine), and coding and compliance specialist roles. He also emphasized the importance of proper training for the various types of sleep disorders that sleep clinics are now treating, evidenced by this pull quote:
“Complex patients need the support from staff during the day to prepare for safe, high quality studies at night. Technical staff needs to be properly trained to ensure everything runs smoothly with HSATs, PAP naps, other versions of mask accommodation for PAP adherence, advanced treatments for ventilator transition and advanced titrations require more daytime staffing than ever.”Anastasi also discussed CPAP compliance in detail, and how each of the different new roles help address PAP compliance. “CPAP is the gold standard treatment. PAP adherence has become a key metric that determines better outcomes.” Sleep educator and sleep navigator roles both address this by acting upon findings in the clinical literature that psychological and educational interventions improve CPAP usage when compared to mechanical interventions alone. Basically, your technologists need more time with patients than ever before. In Anastasi’s opinion, finding ways to provide this type of added time for patient care is critical for the future of sleep medicine and for improving PAP compliance. That’s actually one of the biggest values EnsoSleep brings to your organization. If you’re looking to find more time for patient care, cutting down on time spent scoring is a great place to start. Head over to our Demo Request page to set up a conversation.