Automated Scoring of Sleep and Associated Events

Autonomic arousal detection and cardio-respiratory sleep staging improve the accuracy of home sleep apnea tests

“…To obtain ground-truth scorings following the latest standards, sleep stages, cortical arousals, and respiratory events were scored with the Somnolyzer auto-scoring system embedded as an optional component in Sleepware G3 version 4.0.2.0 for all PSG recordings according to the sleep and sleep disordered breathing scoring rules recommended in the AASM manual version 2.6 ( Berry et al, 2020 ). Somnolyzer has recently been validated and shown to be non-inferior to manual expert scoring regarding sleep stages ( Bakker et al, 2023 ), as well as the apnea-hypopnea index (AHI) and the arousal index (ArI) ( Anderer et al, 2022 ). Therefore, the Somnolyzer auto-scoring system has been cleared by the U.S. Food & Drug administration to be used with adults to generate an output that is ready for review and interpretation by a physician (510K number: K202142).…”

“…Historically, different hypopnea scoring criteria have been defined, requiring flow amplitude reductions of 30 or 50 percent, confirmation by 3% or 4% relative desaturations, confirmation with or without arousals, and sometimes even without any additional confirmation if the amplitude reduction was at least 50 percent ( AASM, 1999 ; Iber et al, 2007 ; Berry et al, 2012 ). The impact of applying different hypopnea scoring rules on the diagnostic outcome has been studied in previous work, suggesting that using arousal events for the scoring of hypopneas facilitates the identification and treatment of a wider spectrum of symptomatic patients without a significant loss of scoring reliability ( Ruehland et al, 2009 ; Berry et al, 2012 ; Anderer et al, 2022 ). Therefore, in 2012, the American Academy of Sleep Medicine (AASM) revised the recommended scoring rule for hypopneas in PSG recordings of adults: a respiratory event shall be scored as hypopnea if there is a 30% reduction in peak signal excursions from the pre-event baseline in the respiratory flow signal for at least 10 s and there is a ≥3% oxygen desaturation from pre-event baseline or the event is associated with an arousal ( Berry et al, 2012 ).…”

Novel Technologies in the Diagnosis and Management of Sleep-disordered Breathing

“…Based on these probabilities, it is possible to create a hypnodensity chart from autoscoring which can be directly compared to the hypnodensity chart based on multiple expert scorings. In Table 2 we summarize publications using AI-algorithms for sleep staging and reporting autoscored hypnodensity graphs (Stephansen et al, 2018;Cesari et al, 2021Cesari et al, , 2022Vallat and Walker, 2021;Anderer et al, 2022b;Brandmayr et al, 2022;Bakker et al, 2023;Fiorillo et al, 2023b). In addition to information regarding the training datasets, the epoch encoder including feature extraction, the sequence encoder and classifier, and the test datasets, the Cohen's kappa values obtained in each study are given.…”

“…A validation study of the Somnolyzer algorithm by Anderer et al (2022b) with 426 PSGs (224 PSGs from the MESA study (Chen et al, 2015), 178 PSGs from the HomePAP study and 24 PSGs from the ABC study (Bakker et al, 2018) scored by one scorer resulted in a Cohen's kappa between Somnolyzer autoscoring and manual scoring of 0.739 (with a 95% confidence interval of 0.737 to 0.741), reflecting substantial agreement according to the thresholds defined by Landis and Koch (1977). In agreement with human inter-rater reliability studies, the highest kappa values were observed for wake and REM detection (0.85 and 0.87) followed by N2 and N3 detection (0.72 and 0.73), while the detection of N1 resulted in the lowest kappa value of 0.46.…”

“…Table 3 summarizes validation results of AI-algorithms that applied a hold-out or crossvalidation; i.e., an internal validation based on data from the same dataset that has been used for training (Supratak et al, 2017;Sors et al, 2018;Phan et al, 2019;Abou Jaoude et al, 2020;Guillot et al, 2020;Korkalainen et al, 2020;Sun et al, 2020a;Alvarez-Estevez and Rijsman, 2021;Fiorillo et al, 2021Fiorillo et al, , 2023bJia et al, 2021;Nasiri and Clifford, 2021;Olesen et al, 2021;Pathak et al, 2021;Vallat and Walker, 2021;Brandmayr et al, 2022;Cho et al, 2022;Ji et al, 2022;Li C. et al, 2022;Sharma et al, 2022;Yubo et al, 2022). Table 4 summarizes the validation results of AI-algorithms which have been validated in datasets completely unseen by the model (Anderer et al, 2018(Anderer et al, , 2022bBiswal et al, 2018;Patanaik et al, 2018;Stephansen et al, 2018;Abou Jaoude et al, 2020;Alvarez-Estevez and Rijsman, 2021;Cesari et al, 2021Cesari et al, , 2022Vallat and Walker, 2021;Bakker et al, 2023). The reported Cohen's kappa values ranged between 0.60 [external validation in 70 patients with Parkinson's disease (Patanaik et al, 2018)] to 0.91 [internal 20-fold epochwise cross validation in 8 healthy subjects (Li C. et al, 2022)].…”