AI Sleep Scoring is the automated event detection and analysis of PSG or HSAT data performed by AI Scoring software, like EnsoSleep. The AI does this by analyzing the waveform data of sleep tests and providing an AI-scored study for a registered clinician to review.
EnsoSleep is FDA-cleared SaMD that automates sleep staging and event detection for patients ages 13 and up. EnsoSleep integrates with existing sleep lab software and devices, automatically detecting studies that require scoring. EnsoSleep analyzes and scores the studies, then uploads the scored studies back into the native software.
Sleep Professionals working at labs with multiple locations and/or devices often prefer to optimize their sleep lab workflow by utilizing EnsoHub to view, edit, and report on all sleep studies in one cloud-based platform.
Diagnostic AHI Severity Agreement
Sleep Disordered Breathing Event Agreement
Sleep Staging Agreement
Central Apnea Detection Agreement
AASM Autoscoring Certification™ is a trademark of the American Academy of Sleep Medicine and is used with permission.
Artificial Intelligence (AI) is the general field of study at the intersection of statistics, software development and theoretical computer science.
AI developers ideate different algorithmic approaches using large data sets to create machines with a certain level of intelligence.
Machine Learning (ML) is a subset of AI, and refers to a collection of methods that enable an algorithm to execute specific tasks without being explicitly programmed.
An ML algorithm learns relationships within the observed data in order to make predictions on newly collected data.
Auto scoring solutions are developed using traditional software-assisted programming, where humans manually “code” or “write” individual rules that are applied to sleep study data to generate answers. The number of rules that have to be coded to account for the high level of variability in sleep data makes traditional software-assisted auto scoring approaches unsuitable for sleep studies. Instead of providing time savings, sleep technologists and clinicians often spend more time correcting errors caused by the auto scoring.
In contrast to auto scoring, EnsoSleep uses a subset of AI called Machine Learning (ML). ML solutions do not require humans to individually code all the rules. Instead, ML software learns relationships from >2.8 million RPSGT-scored and reviewed sleep studies (“answers”) and applies AASM scoring guidelines to develop methods for the analysis and scoring of studies. EnsoData’s AI/ML algorithms are regularly maintained, tested, and validated over time, so every sleep study analyzed helps improve predictions on newly collected data.
EnsoData’s cloud database for AI and ML development and testing includes more than 2,800,000 sleep studies across age, gender, race, and other demographic categories. This data includes more than 10,000,000 hours of second-to-second physiological signal data that our AI and ML algorithms can learn from. EnsoData’s AI models are then trained on the rules established by the “AASM gold standard” panel of sleep physicians.
EnsoData’s™ AI-powered technology is built on an ethically sound foundation and uses a transparent process for design, validation, and improvement of AI systems to mitigate risks of bias. Before each minor software release, EnsoData’s engineering team assesses updates to the models for bias using the train and test method. Furthermore, the algorithm only processes signal data and does not have access to demographic information found in a sleep study.
As studies are processed by EnsoSleep, our proprietary AI algorithm progressively learns sleep technologists’ scoring tendencies by analyzing the study data. This enables the AI to make more informed decisions on future sleep studies, ensuring the highest level of initial scoring accuracy, and saving your sleep team time by eliminating repeated edits.
For example, post-arousal sighs are often categorized as an event and most techs choose to delete them. EnsoSleep allows for setting a specific duration after arousal, during which no events are scored. This effectively eliminates the need for extra deletion of post-arousal sighs.
In addition to the consistency the EnsoSleep algorithm provides, customers using EnsoSleep can also personalize their scoring settings to match the unique preferences of the sleep lab, resulting in improved workflow efficiency and optimization.
EnsoSleep scores all sleep studies in accordance with AASM standards; however, it is our experience that most sleep clinics have unique preferences specific to their practice. The following settings can be personalized to match how studies are currently scored in your sleep lab.
Yes, EnsoSleep scores home sleep apnea tests (HSATs). In fact, AI scoring makes at-home testing more scalable and profitable for providers as it simultaneously improves consistency and reduces time spent scoring.
Customers using EnsoSleep AI scoring to take the first pass at scoring studies results in an average 62% time savings on PSGs and an average 68% time savings on HSATs.
This provides clinicians with more time to expand patient care, improve outcomes, and treat more patients. Click the button below to view the time savings survey results.
EnsoSleep works behind the scenes, integrating with leading PSG and HSAT software and devices. Working directly with your software, EnsoSleep automatically detects studies that need scoring, analyzes them, and places them back into the existing scoring workflow.
Yes. EnsoHub provides clinicians with access to ALL studies and devices, across ALL locations, for ALL sleep techs and physicians, in one interoperable, cloud-based platform for viewing, editing, and reporting of sleep studies.
Schedule a demo to explore how EnsoSleep streamlines diagnosis, supports sleep lab growth, and elevates the patient experience.