Pith. sign in

REVIEW 3 major objections 5 minor 53 references

Apnea Burden-Guided Framework: Enhancing Out-of-Distribution Generalization in PPG-Based Sleep Apnea Characterization

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that adding PPG pulse-wave morphology features to SpO2, inside a hybrid convolutional-recurrent network that first predicts apnea burden, improves sleep apnea severity classification and markedly improves…

desk verdict Solid comparison study with a plausible direction, but the headline OOD gain is not yet separated from a shaky AB-to-AHI conversion and 30-subject statistics. read the letter →

arxiv 2608.12229 v1 pith:IKB6DJAG submitted 2026-08-12 eess.SP

Mantas Rinkevičius , Oskar Pfeffer , Amal Alissa , Nando Hegemann , Vaidotas Marozas This is my paper · ORCID
classification eess.SP
keywords apneaburdenout-of-distributiongeneralizationphotoplethysmographyPPGpulsewavefeaturessleepseveritySpO2convolutional-recurrentnetworksapnea-hypopneaindex
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to establish that the shape of the PPG pulse wave carries information about sleep apnea that oxygen saturation alone misses, and that adding this information to a neural network makes severity grading more robust when the model meets patients from a different population. The proposed pipeline first predicts apnea burden, the fraction of time spent in respiratory events, as a continuous number, then converts it into the clinical apnea-hypopnea index using a fixed mean event duration from training data, and assigns one of four severity classes. On an external cohort of 30 stroke patients, the best hybrid model's macro-F1 rose from 28.82% with SpO2 alone to 44.83% with six PPG features plus SpO2, while macro-sensitivity rose from 31.90% to 47.62%. This matters because wearable home monitoring already records both PPG and SpO2, and robustness to a new device or population is the main barrier to practical, long-term screening.

What carries the argument

The load-bearing objects are: the apnea burden target, defined as the fraction of a 5-minute segment occupied by apnea or hypopnea events, which turns the problem into regression rather than classification; the conversion $AHI_{AB} = AB \cdot 3600 / D_{mean}$ with the single training-set constant $D_{mean} = 21.39$ s; and the predictor set, five standard pulse wave features (pulse wave interval $T_p$, peak-to-peak amplitude $A_{pp}$, systolic time $ST$, diastolic time $DT$, and maximum slope $S_{max}$) plus the proposed mean envelope $E(T_p)$, the average of the upper and lower envelopes of the pulse-to-pulse interval time series within a 1-minute epoch. The argument's force is that predicting burden first and thresholding the converted AHI, with hybrid convolutional-recurrent models, transfers better to an unseen stroke-patient cohort than classifying directly from SpO2 alone.

What would settle it

Re-run the trained models on the OSASUD cohort and recompute AHI from the predicted apnea burden using the cohort's own mean event duration instead of 21.39 s; if several subjects move across the 15 or 30 counts/h severity thresholds, the fixed-duration conversion is carrying the result. As a second check, test the same PPG-plus-SpO2 fusion on another independent external dataset: if the macro-F1 gain relative to SpO2 alone disappears, the claimed out-of-distribution benefit is specific to the 30-subject stroke cohort studied here.

Watch

Extended reading notes

Core claim

The paper's central claim, stated in its own conclusion, is that fusing PPG-derived pulse wave features with SpO2 in artificial neural networks improves sleep apnea severity classification and enhances out-of-distribution generalization. The authors argue that a regression-based pipeline that predicts continuous apnea burden per 5-minute segment, aggregates the predictions per subject, and converts them into AHI via $AB \cdot 3600 / D_{mean}$, then thresholds into normal, mild, moderate, and severe, is a clinically aligned alternative to direct classification. Averaged over five seeds, the 2CNN-2LSTM model using PPG features, SpO2, and the proposed mean-envelope feature $E(T_p)$ achieved the highest overall performance, with OOD macro-accuracy 65.36%, macro-F1 44.83%, and macro-sensitivity 47.62%, versus 56.14%, 28.82%, and 31.90% for SpO2 alone. The authors further claim that the newly proposed PPG features, particularly the mean envelope of the pulse wave interval, add a modest but consistent benefit under distribution shift.

Load-bearing premise

The entire AHI estimate and severity label rest on a single constant $D_{mean} = 21.39$ s, the mean apnea/hypopnea event duration computed from the training set and applied to every subject, although event durations vary from about 10 seconds to 2 minutes; if that constant is unrepresentative for a population, every AHI value and every severity boundary shifts.

Editorial extensions

If this is right

  • A wearable that records PPG and SpO2 could assign a patient to one of the four clinical severity categories without full polysomnography, using only the burden-to-AHI conversion.
  • PPG pulse wave morphology contributes the largest performance gains precisely when the test population differs from training, which is the situation home monitoring devices actually face.
  • Low-complexity hybrid convolutional-recurrent models (2CNN-GRU and 2CNN-2LSTM) generalize better than a larger multi-scale convolutional model, suggesting that model simplicity is not a barrier to out-of-distribution robustness.
  • The proposed mean-envelope feature $E(T_p)$ adds a small but consistent benefit for hybrid architectures, supporting the view that minute-scale autonomic modulation is informative for apnea severity.
  • Regression toward apnea burden, followed by conversion to AHI, aligns model output with clinical severity thresholds while avoiding direct classification decisions at the segment level.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The paper leaves untested whether the same PPG-plus-SpO2 gain appears on a second independent dataset with different devices and demographics; that experiment would separate a general out-of-distribution benefit from one specific to the 30-subject stroke cohort.
  • The paper does not quantify how sensitive its results are to the fixed $D_{mean}$ conversion, although it flags the risk; a sensitivity analysis that perturbs $D_{mean}$ across the observed 10-second to 2-minute duration range would show how many borderline patients change severity class.
  • If the mean-envelope feature really captures autonomic arousal during apnea, it may transfer to other autonomic-modulation screening tasks, such as nocturnal arrhythmia or stress monitoring; that is an untested implication of the proposed physiology.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper proposes an apnea burden (AB) guided framework for sleep apnea severity classification from PPG and SpO2 signals. Segment-level AB is predicted by three neural network architectures (2CNN-GRU, 2CNN-2LSTM, and parallel CNNs), aggregated to subject level, and converted to AHI via a fixed mean event duration D_mean = 21.39 s estimated from the MESA training data. The framework is evaluated on MESA as in-distribution data and on OSASUD (30 stroke patients) as out-of-distribution data. The headline claim is that adding PPG-derived features (including a new mean-envelope feature E(Tp)) to SpO2 improves OOD macro-F1 from 28.82% to 44.83% for the 2CNN-2LSTM architecture.

Significance. If the results are robust, the paper offers a practical, wearable-compatible pipeline for sleep apnea severity assessment that explicitly addresses out-of-distribution generalization. The study is reasonably designed, with two external cohorts, multiple architectures, and five repeated runs per configuration. However, the central OOD improvement rests on a small 30-subject cohort and on an AB-to-AHI conversion that may introduce unquantified bias, making the headline gains less certain than the text suggests.

major comments (3)
  1. [II-E, Equation (4) and Figure 8] The validation of the AB-to-AHI conversion is circular. Because AB is computed from the same annotated events and D_mean is derived from the training data, the near-identity relationship in Figure 8 is an algebraic consequence of the definitions, not independent empirical validation. The reported MAE between AHI_AB and reference AHI therefore does not validate the conversion for out-of-distribution cohorts. The authors should either present a non-circular justification for the fixed D_mean or quantify how classification outcomes change under plausible D_mean values for OSASUD.
  2. [III.C and Table 4] The OOD evaluation uses only 30 subjects and no statistical inference. The standard deviations reported in Table 4 reflect five training runs, not subject-level sampling variability. For example, the 2CNN-2LSTM OOD macro-F1 differences (28.82±6.18 vs. 44.83±6.94) could be compatible with considerable overlap under subject-level bootstrap resampling. I request subject-level bootstrap confidence intervals or a paired significance test (e.g., McNemar or bootstrap for macro-F1) for the key SpO2-only vs. PPG+SpO2 comparisons.
  3. [III.A and IV.C] The sensitivity of the severity classification to the choice of D_mean is not analyzed, and this is load-bearing because the severity boundaries (5, 15, 30) are applied to AHI_AB. The paper acknowledges this in Section IV.C but does not quantify it. I ask the authors to report the mean event duration in the OSASUD test set from the annotated labels, compare it to D_mean = 21.39 s, and re-evaluate the OOD headline metrics with D_mean varied across the plausible range (e.g., ±30%). This will show whether the reported PPG-vs-SpO2 improvements are robust to the conversion constant.
minor comments (5)
  1. [II-C] The description of E(Tp) in Equation (1) is brief; please specify the window length and the exact interpolation procedure more precisely, as the implementation is not fully reproducible from the text.
  2. [III.A and Equation (4)] D_mean is first used in Section II-E but its definition (mean duration of all annotated events in the training set) appears only in Section III.A. Please define it earlier and state why a fixed value is clinically acceptable.
  3. [Figure 3 and Table 4] The severity labels 'normal' and 'no apnea' are used interchangeably. Please unify the terminology to avoid confusion.
  4. [II-B] The sentence 'a zero-phase, fourth-order Butterworth band-pass filter (0.4–9 Hz)' should clarify whether the filter is applied before or after resampling; as written it is ambiguous.
  5. [II-C] The phrase 'We assume that the maximum slope of PPG, Smax, may provide complementary information' is speculative; consider rephrasing as a hypothesis or expectation.

Circularity Check

1 steps flagged · score 5.0 of 10

AB-to-AHI conversion is a definitional rescaling, so Figure 8 is not independent validation; the PPG-vs-SpO2 OOD comparison itself is genuine.

  1. self definitional [Section II.E, Eq. (4) and Section III.A / Figure 8]
    "The predicted AHI is computed as: AHIAB = AB·3600 / Dmean, where AB denotes the overall predicted AB, and Dmean represents the mean duration of apnea/hypopnea events (in seconds) calculated from overnight recordings of the training data. ... the scatter plots show that AHIAB closely follows the reference AHI values along the identity line (y=x), indicating that AB provides a consistent surrogate for event frequency when normalized by the average event duration."

    AB is defined in Eqs. (2)-(3) as the total apnea/hypopnea duration divided by recording time, while AHI is the number of events per hour. For any subject, AHI = AB * 3600 / D_true exactly, where D_true is that subject's mean event duration. Eq. (4) replaces D_true with the training-set constant D_mean = 21.39 s. Therefore AHI_AB is an algebraic rescaling of AB, and Figure 8's near-identity scatter is not an empirical confirmation that AB is a 'consistent surrogate'; it merely shows how well one global mean duration approximates each subject's own mean duration. The conversion is definitionally circular as a validation step, although the model's AB predictions themselves are empirical.

full rationale

The central empirical claim—that adding PPG-derived features to SpO2 improves OOD sleep-apnea severity classification—is supported by genuinely fitting segment-level AB regressors on MESA and testing on OSASUD. That comparison is not manufactured by Eq. (4): all predictor combinations share the same fixed D_mean conversion, and the reported macro-F1, kappa, and MCC gains arise from the trained models, not from the conversion constant. The circularity is confined to the paper's presentation of the AB-to-AHI mapping: because AHI and AB are tied by the definitional identity AHI = AB * 3600 / D_true, the 'estimated' AHI_AB is a rescaled AB, and Figure 8's identity-line agreement is a restatement of the definitions plus the variability of event durations, not an independent validation. The paper itself acknowledges this in Section IV.C, noting that a fixed average duration may introduce bias across populations. No load-bearing self-citation was found; the authors' prior MESA-related citations are contextual rather than justificatory, and no uniqueness theorem or ansatz is smuggled in via citation. Overall, the feature-ablation conclusion is independent, but the conversion validation and the AHI_AB endpoint contain a partial definitional circularity, so the paper receives a moderate score rather than a clean bill.

Assumptions & free parameters 1 free parameters · 4 assumptions · 1 invented entities

The central empirical claim rests on the AB-to-AHI conversion using a training-set constant, on the physiological relevance of PPG features, and on the representativeness of the two cohorts. These are domain assumptions rather than derived statements.

free parameters (1)
  • D_mean = 21.39 s
    Mean duration of apnea/hypopnea events computed from the MESA training subset and used in Equation (4) to convert every subject's AB to AHI. If event durations differ across populations, AHI estimates are systematically biased.
assumptions (4)
  • ad hoc to paper The fixed mean event duration D_mean from MESA training data applies to OOD subjects for the AB-to-AHI conversion.
    Equation (4) uses D_mean = 21.39 s for all conversions; Section IV.C acknowledges event durations vary from about 10 s to 2 min and may bias AHI estimates.
  • domain assumption AB is a clinically meaningful continuous surrogate for AHI.
    AB and AHI are computed from the same event annotations; the paper assumes temporal extent of events is linearly related to event frequency through mean duration (Section II.E).
  • domain assumption PPG morphological features capture apnea-related autonomic and hemodynamic information beyond SpO2.
    This is the tested hypothesis, supported by cited prior work [12,13,14,32]; the paper does not independently verify the physiological mechanism.
  • domain assumption The MESA quality-grade-7 subset represents the ID distribution and OSASUD represents a valid OOD population.
    Only 276 of 2055 MESA subjects are selected by quality grade, and OSASUD has only 30 stroke patients, so the OOD evidence is preliminary (Sections II.A and IV.C).
invented entities (1)
  • E(Tp), the mean envelope of the pulse wave interval
    purpose: Proposed PPG feature intended to capture low-frequency autonomic modulation of heart rate during apnea episodes
    Defined and evaluated only within this study; results show modest and sometimes unstable benefit (Section IV.C), with no external validation of the feature.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Apnea Burden-Guided Framework: Enhancing Out-of-Distribution Generalization in PPG-Based Sleep Apnea Characterization." pith.science (2026). https://pith.science/paper/IKB6DJAG

@misc{pith2026260812229,
  author       = {Pith},
  title        = {Pith review of: Apnea Burden-Guided Framework: Enhancing Out-of-Distribution Generalization in PPG-Based Sleep Apnea Characterization},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IKB6DJAG}},
  note         = {Machine review of arXiv:2608.12229}
}
read the original abstract

Sleep apnea is a common sleep-related breathing disorder associated with substantial cardiovascular and metabolic risk. Although overnight polysomnography remains the reference standard for diagnosis, its complexity and cost limit its suitability for long-term preventive monitoring at home. In wearable systems, arterial blood oxygen saturation (SpO2) is commonly used as the main predictor, whereas additional morphological features of the photoplethysmographic (PPG) pulse wave are usually underexplored. This study proposes a novel apnea burden prediction-based framework for sleep apnea severity assessment and investigates the influence of PPG features on model performance and out-of-distribution (OOD) generalization. The proposed framework first predicts apnea burden as a continuous measure, which is subsequently converted into the clinically relevant apnea-hypopnea index for subject-level classification into four severity groups in both in-distribution (ID) and OOD data. Three artificial neural network architectures were evaluated, and the performance metrics were averaged over five independent runs with different fixed random seeds. During OOD testing, the combination of PPG features and SpO2 led to increases of approximately 15.72% in macro-sensitivity, 9.22% in macro-accuracy, 16.01% in macro-F1-score, 11.08% in Cohen's kappa, and 13.22% in Matthews correlation coefficient, compared with using SpO2 alone. The low-complexity convolutional-recurrent models achieved the highest overall performance. The results indicated that the proposed apnea burden-guided framework, combined with PPG features and SpO2, improves sleep apnea characterization while showing encouraging OOD performance on an independent external dataset. Moreover, simpler hybrid architectures demonstrated strong potential for robust home-based preventive monitoring.

Figures

Figures reproduced from arXiv: 2608.12229 by the authors.

Figure 1
Figure 1. FIGURE 1 [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. FIGURE 2 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. FIGURE 3 [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (9 more)
Figure 5
Figure 5. Figure 5: FIGURE 5 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: FIGURE 6 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: FIGURE 7 [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: FIGURE 8 [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: FIGURE 9 [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: FIGURE 10 [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 11
Figure 11. Figure 11: FIGURE 11 [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]
Figure 12
Figure 12. Figure 12: FIGURE 12 [PITH_FULL_IMAGE:figures/full_fig_p012_12.png]
Figure 13
Figure 13. Figure 13: FIGURE 13 [PITH_FULL_IMAGE:figures/full_fig_p013_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 30 canonical work pages

  1. [1]

    Sleep apnea and cardiovascular disease,

    V . K. Somers et al., “Sleep apnea and cardiovascular disease,”J. Am. Coll. Cardiol., vol. 52, no. 8, pp. 686–717, Aug. 2008, doi: 10.1016/j.jacc.2008.05.002

    work page doi:10.1016/j.jacc.2008.05.002 2008

  2. [2]

    Sleep patterns and arrhythmias: should this keep us awake at night?,

    A. H. Kadish and J. T. Jacobson, “Sleep patterns and arrhythmias: should this keep us awake at night?,”J. Am. Coll. Cardiol., vol. 78, no. 12, pp. 1208–1209, Sep. 2021, doi: 10.1016/j.jacc.2021.07.024

    work page doi:10.1016/j.jacc.2021.07.024 2021

  3. [3]

    Healthy sleep patterns and risk of incident arrhythmias,

    X. Li et al., “Healthy sleep patterns and risk of incident arrhythmias,” J. Am. Coll. Cardiol., vol. 78, no. 12, pp. 1197–1207, Sep. 2021, doi: 10.1016/j.jacc.2021.07.023

    work page doi:10.1016/j.jacc.2021.07.023 2021

  4. [4]

    Obstructive sleep apnoea,

    A. Malhotra and D. P . White, “Obstructive sleep apnoea,”Lancet, vol. 360, no. 9328, pp. 237–245, Jul. 2002, doi: 10.1016/S0140-6736(02)09464-3

    work page doi:10.1016/s0140-6736(02)09464-3 2002

  5. [5]

    Epidemiology of obstructive sleep apnea: a population health perspective,

    T. Y oung, P . E. Peppard, and D. J. Gottlieb, “Epidemiology of obstructive sleep apnea: a population health perspective,”Am. J. Respir . Crit. Care Med., vol. 165, no. 9, pp. 1217–1239, 2002, doi: 10.1164/rccm.2109080

    work page doi:10.1164/rccm.2109080 2002

  6. [6]

    On the relationship between diabetes and obstructive sleep apnea: evolution and epi- genetics,

    N. Wilson, O. J. V eatch, and S. Johnson, “On the relationship between diabetes and obstructive sleep apnea: evolution and epi- genetics,”Biomedicines, vol. 10, no. 3, p. 668, Mar. 2022, doi: 10.3390/biomedicines10030668

    work page doi:10.3390/biomedicines10030668 2022

  7. [7]

    Sleep apnea: relationship to age, sex, and Alzheimer’s dementia,

    R. G. Smallwood, M. V . Vitiello, E. C. Giblin, and P . N. Prinz, “Sleep apnea: relationship to age, sex, and Alzheimer’s dementia,”Sleep, vol. 6, no. 1, pp. 16–22, Sep. 1983, doi: 10.1093/sleep/6.1.16

    work page doi:10.1093/sleep/6.1.16 1983

  8. [8]

    Sleep-related breathing disorders in adults: recommendations for syn- drome definition and measurement techniques in clinical research,

    “Sleep-related breathing disorders in adults: recommendations for syn- drome definition and measurement techniques in clinical research,”Sleep, vol. 22, no. 5, pp. 667–689, Aug. 1999, doi: 10.1093/sleep/22.5.667

    work page doi:10.1093/sleep/22.5.667 1999

Show all 53 references
  1. [9]

    Photoplethysmography and its application in clinical physiolog- ical measurement,

    J. Allen, “Photoplethysmography and its application in clinical physiolog- ical measurement,”Physiol. Meas., vol. 28, no. 3, pp. R1–R39, Feb. 2007, doi: 10.1088/0967-3334/28/3/R01. 14 VOLUME 11, 2023 Rinkevičius et al.: AB-Guided Framework: Enhancing OOD Generalization in PPG-...

    2007 doi

  2. [10]

    On the analysis of fingertip photoplethysmogram sig- nals,

    M. Elgendi, “On the analysis of fingertip photoplethysmogram sig- nals,”Curr . Cardiol. Rev., vol. 8, no. 1, pp. 14–25, Jun. 2012, doi: 10.2174/157340312801215782

    2012 doi

  3. [11]

    Diagnosis of ob- structive sleep apnea using pulse oximeter derived photoplethysmographic signals,

    A. Romem, A. Romem, D. Koldobskiy, and S. M. Scharf, “Diagnosis of ob- structive sleep apnea using pulse oximeter derived photoplethysmographic signals,”J. Clin. Sleep Med., vol. 10, no. 3, pp. 285–290, Mar. 2014, doi: 10.5664/JCSM.3530

    2014 doi

  4. [12]

    Pulse rate variability analysis for discrimination of sleep-apnea-related decreases in the ampli- tude fluctuations of pulse photoplethysmographic signal in children,

    J. Lázaro, E. Gil, J. M. V ergara, and P . Laguna, “Pulse rate variability analysis for discrimination of sleep-apnea-related decreases in the ampli- tude fluctuations of pulse photoplethysmographic signal in children,”IEEE J. Biomed. Health Inform., vol. 18, no. 1, pp. 240–24...

    2014

  5. [13]

    Detection and Classification of Sleep Apnea and Hypopnea Using PPG and SpO2 Signals,

    R. Lazazzera et al., “Detection and Classification of Sleep Apnea and Hypopnea Using PPG and SpO2 Signals,”IEEE Transactions on Biomedical Engineering, vol. 68, no. 5, pp. 1496–1506, May 2021, doi: 10.1109/tbme.2020.3028041

    2021

  6. [14]

    Detection of decreases in the amplitude fluctuation of pulse photoplethysmography signal as in- dication of obstructive sleep apnea syndrome in children,

    E. Gil, J. María V ergara, and P . Laguna, “Detection of decreases in the amplitude fluctuation of pulse photoplethysmography signal as in- dication of obstructive sleep apnea syndrome in children,”Biomed. Signal Process. Control, vol. 3, no. 3, pp. 267–277, Jul. 2008, doi: 10...

    2008 doi

  7. [15]

    Analysis of differential photoplethysmography signal patterns in apnea and hypopnea,

    M. Á. Goda, A. Oksenberg, A. Azarbarzin, and J. A. Behar, “Analysis of differential photoplethysmography signal patterns in apnea and hypopnea,” Physiol. Meas., vol. 47, no. 2, p. 025006, Feb. 2026, doi: 10.1088/1361- 6579/ae3ef0

    2026 doi

  8. [16]

    Deep learning,

    Y . LeCun, Y . Bengio, and G. Hinton, “Deep learning,”Nature, vol. 521, no. 7553, pp. 436–444, May 2015, doi: 10.1038/nature14539

    2015 doi

  9. [17]

    Deep learning for healthcare applications based on physiological signals: A review,

    O. Faust, Y . Hagiwara, T. J. Hong, O. S. Lih, and U. R. Acharya, “Deep learning for healthcare applications based on physiological signals: A review,”Comput. Methods Programs Biomed., vol. 161, pp. 1–13, Jul. 2018, doi: 10.1016/j.cmpb.2018.04.005

    2018 doi

  10. [18]

    A systematic review of detecting sleep apnea using deep learning,

    S. S. Mostafa, F. Mendonça, A. G. Ravelo-García, and F. Morgado-Dias, “A systematic review of detecting sleep apnea using deep learning,”Sensors, vol. 19, no. 22, p. 4934, Nov. 2019, doi: 10.3390/s19224934

    2019 doi

  11. [19]

    Deep- learning based sleep apnea detection using SpO2 and pulse rate,

    P . Sharma, A. Jalali, M. Majmudar, K. S. Rajput, and N. Selvaraj, “Deep- learning based sleep apnea detection using SpO2 and pulse rate,” inProc. 44th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. (EMBC), Jul. 2022, doi: 10.1109/EMBC48229.2022.9871295

    2022

  12. [20]

    Deep-learning based sleep apnea detection using sleep sound, SpO2, and pulse rate,

    C. Singtothong and T. Siriborvornratanakul, “Deep-learning based sleep apnea detection using sleep sound, SpO2, and pulse rate,”Int. J. Inf. Technol., vol. 16, no. 8, pp. 4869–4874, May 2024, doi: 10.1007/s41870- 024-01906-x

    2024 doi

  13. [21]

    DeepSleepNet: A model for automatic sleep stage scoring based on raw single-channel EEG,

    A. Supratak, H. Dong, C. Wu, and Y . Guo, “DeepSleepNet: A model for automatic sleep stage scoring based on raw single-channel EEG,”IEEE Trans. Neural Syst. Rehabil. Eng., vol. 25, no. 11, pp. 1998–2008, Nov. 2017, doi: 10.1109/TNSRE.2017.2721116

    1998

  14. [22]

    Ap- neaNet: A hybrid 1D-CNN-LSTM architecture for detection of obstructive sleep apnea using digitized ECG signals,

    G. Srivastava, A. Chauhan, N. Kargeti, N. Pradhan, and V . S. Dhaka, “Ap- neaNet: A hybrid 1D-CNN-LSTM architecture for detection of obstructive sleep apnea using digitized ECG signals,”Biomed. Signal Process. Control, vol. 84, p. 104754, Jul. 2023, doi: 10.1016/j.bspc.2023.104754

    2023

  15. [23]

    Transformer-based deep learning approach for obstructive sleep apnea detection using single-lead ECG,

    M. A. Almarshad, S. Al-Ahmadi, S. Islam, A. Soudani, and A. S. BaHam- mam, “Transformer-based deep learning approach for obstructive sleep apnea detection using single-lead ECG,”Front. Artif. Intell., vol. 9, Feb. 2026, doi: 10.3389/frai.2026.1727091

    2026

  16. [24]

    Advancing sleep disorder diagnostics: A transformer-based EEG model for sleep stage classification and OSA prediction,

    C. Wan, M. C. Nnamdi, W. Shi, B. Smith, C. Purnell, and M. D. Wang, “Advancing sleep disorder diagnostics: A transformer-based EEG model for sleep stage classification and OSA prediction,”IEEE J. Biomed. Health Inform., pp. 1–9, 2024, doi: 10.1109/JBHI.2024.3512616

    2024

  17. [25]

    Deep learning for obstructive sleep apnea detection and severity assessment: A multimodal signals fusion multiscale trans- former model,

    Y . Zhang et al., “Deep learning for obstructive sleep apnea detection and severity assessment: A multimodal signals fusion multiscale trans- former model,”Nat. Sci. Sleep, vol. 17, pp. 1–15, Jan. 2025, doi: 10.2147/NSS.S492806

    2025 doi

  18. [26]

    Adoption of transformer neural network to improve the diagnostic performance of oximetry for obstructive sleep apnea,

    M. A. Almarshad, S. Al-Ahmadi, M. S. Islam, A. S. BaHammam, and A. Soudani, “Adoption of transformer neural network to improve the diagnostic performance of oximetry for obstructive sleep apnea,”Sensors, vol. 23, no. 18, p. 7924, Sep. 2023, doi: 10.3390/s23187924

    2023 doi

  19. [27]

    SomnNET: An SpO2- based deep learning network for sleep apnea detection in smartwatches,

    A. John, K. K. Nundy, B. Cardiff, and D. John, “SomnNET: An SpO2- based deep learning network for sleep apnea detection in smartwatches,” inProc. Annu. Int. Conf. IEEE Eng. Med. Biol. Soc. (EMBC), Nov. 2021, pp. 1961–1964, doi: 10.1109/EMBC46164.2021.9631037

    2021

  20. [28]

    SpO2 based sleep apnea detection using deep learning,

    S. S. Mostafa, F. Mendonça, F. Morgado-Dias, and A. G. Ravelo-García, “SpO2 based sleep apnea detection using deep learning,” inProc. Int. Conf. Intell. Eng. Syst. (INES), Oct. 2017, doi: 10.1109/INES.2017.8118534

    2017

  21. [29]

    Deep learn- ing approaches for assessing pediatric sleep apnea severity through SpO2 signals,

    E. Mortazavi, B. Tarvirdizadeh, K. Alipour, and M. Ghamari, “Deep learn- ing approaches for assessing pediatric sleep apnea severity through SpO2 signals,”Sci. Rep., vol. 14, no. 1, Oct. 2024, doi: 10.1038/s41598-024- 67729-9

    2024 doi

  22. [30]

    AI-driven sleep apnea screening with overnight blood oxygen saturation: current practices and future directions,

    N. H. Hoang and Z. Liang, “AI-driven sleep apnea screening with overnight blood oxygen saturation: current practices and future directions,”Front. Digit. Health, vol. 7, Apr. 2025, doi: 10.3389/fdgth.2025.1510166

    2025

  23. [31]

    Automatic classification of apnea/hypopnea events through sleep/wake states and severity of SDB from a pulse oximeter,

    J. U. Park, H. K. Lee, J. Lee, E. Urtnasan, H. Kim, and K. J. Lee, “Automatic classification of apnea/hypopnea events through sleep/wake states and severity of SDB from a pulse oximeter,”Physiol. Meas., vol. 36, no. 9, pp. 2009–2025, Aug. 2015, doi: 10.1088/0967-3334/36/9/2009

    2009 doi

  24. [32]

    ApSense: Data-driven algorithm in PPG-based sleep apnea sensing,

    T. Choksatchawathi et al., “ApSense: Data-driven algorithm in PPG-based sleep apnea sensing,”IEEE Internet Things J., vol. 11, no. 20, pp. 33915– 33926, Oct. 2024, doi: 10.1109/JIOT.2024.3433500

    2024

  25. [33]

    SleepPPG-Net2: Deep learning generalization for sleep staging from photoplethysmography,

    S. Attia et al., “SleepPPG-Net2: Deep learning generalization for sleep staging from photoplethysmography,”Physiol. Meas., vol. 46, no. 12, p. 125001, Dec. 2025, doi: 10.1088/1361-6579/ae1a34

    2025 doi

  26. [34]

    Racial/ethnic differences in sleep disturbances: The Multi- Ethnic Study of Atherosclerosis (MESA),

    X. Chen et al., “Racial/ethnic differences in sleep disturbances: The Multi- Ethnic Study of Atherosclerosis (MESA),”Sleep, vol. 38, no. 6, Jun. 2015, doi: 10.5665/sleep.4732

    2015 doi

  27. [35]

    Obstruc- tive sleep apnea event detection using explainable deep learning mod- els for a portable monitor,

    Á. S. Alarcón, N. M. Madrid, R. Seepold, and J. A. Ortega, “Obstruc- tive sleep apnea event detection using explainable deep learning mod- els for a portable monitor,”Front. Neurosci., vol. 17, Jul. 2023, doi: 10.3389/fnins.2023.1155900

    2023

  28. [36]

    Obstructive sleep apnea characterization: A mul- timodal cross-recurrence-based approach for investigating atrial fibrilla- tion,

    M. Rinkevičius et al., “Obstructive sleep apnea characterization: A mul- timodal cross-recurrence-based approach for investigating atrial fibrilla- tion,”IEEE J. Biomed. Health Inform., vol. 28, no. 10, pp. 6155–6167, Jul. 2024, doi: 10.1109/JBHI.2024.3428845

    2024

  29. [37]

    Influ- ence of photoplethysmogram signal quality on pulse arrival time dur- ing polysomnography,

    M. Rinkevičius, P . H. Charlton, R. Bailón, and V . Marozas, “Influ- ence of photoplethysmogram signal quality on pulse arrival time dur- ing polysomnography,”Sensors, vol. 23, no. 4, p. 2220, Feb. 2023, doi: 10.3390/s23042220

    2023 doi

  30. [38]

    OSASUD: A dataset of stroke unit recordings for the detection of ob- structive sleep apnea syndrome,

    A. Bernardini, A. Brunello, G. L. Gigli, A. Montanari, and N. Saccomanno, “OSASUD: A dataset of stroke unit recordings for the detection of ob- structive sleep apnea syndrome,”Sci. Data, vol. 9, no. 1, Apr. 2022, doi: 10.1038/s41597-022-01272-y

    2022 doi

  31. [39]

    AIOSA: An approach to the automatic identification of obstructive sleep apnea events based on deep learning,

    A. Bernardini, A. Brunello, G. L. Gigli, A. Montanari, and N. Saccomanno, “AIOSA: An approach to the automatic identification of obstructive sleep apnea events based on deep learning,”Artif. Intell. Med., vol. 118, p. 102133, Aug. 2021, doi: 10.1016/j.artmed.2021.102133

    2021

  32. [40]

    Accurate apnea and hypopnea localization in PSG with multi-scale object detection via dual-modal fea- ture learning,

    Y . Ji, D. Chen, Y . Zuo, T. Gao, and Y . Tang, “Accurate apnea and hypopnea localization in PSG with multi-scale object detection via dual-modal fea- ture learning,”Biomed. Signal Process. Control, vol. 89, p. 105717, Nov. 2023, doi: 10.1016/j.bspc.2023.105717

    2023

  33. [41]

    Obstructive sleep apnea syndrome identification using CNN-LSTM hybrid model,

    P . Kulkarni et al., “Obstructive sleep apnea syndrome identification using CNN-LSTM hybrid model,”J. Electr . Syst., vol. 20, no. 2, pp. 2386–2394, Apr. 2024, doi: 10.52783/jes.2013

    2024 doi

  34. [42]

    Infor- mation retrieval from photoplethysmographic sensors: a comprehensive comparison of practical interpolation and breath-extraction techniques at different sampling rates,

    P . Reali, R. Lolatto, S. Coelli, G. Tartaglia, and A. M. Bianchi, “Infor- mation retrieval from photoplethysmographic sensors: a comprehensive comparison of practical interpolation and breath-extraction techniques at different sampling rates,”Sensors, vol. 22, no. 4, p. 1428,...

    2022 doi

  35. [43]

    An automatic beat detection algorithm for pressure signals,

    M. Aboy, J. McNames, T. Thong, D. Tsunami, M. S. Ellenby, and B. Goldstein, “An automatic beat detection algorithm for pressure signals,” IEEE Trans. Biomed. Eng., vol. 52, no. 10, pp. 1662–1670, Oct. 2005, doi: 10.1109/TBME.2005.855725

    2005

  36. [44]

    PP-Net: A deep learning framework for PPG-based blood pressure and heart rate estima- tion,

    M. Panwar, A. Gautam, D. Biswas, and A. Acharyya, “PP-Net: A deep learning framework for PPG-based blood pressure and heart rate estima- tion,”IEEE Sens. J., vol. 20, no. 17, pp. 10000–10011, Sep. 2020, doi: 10.1109/JSEN.2020.2990864

    2020

  37. [45]

    Going deeper with convolutions,

    C. Szegedy et al., “Going deeper with convolutions,” inProc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2015, pp. 1–9, doi:n10.1109/cvpr.2015.7298594

    2015

  38. [46]

    Metrics for multi-class classifica- tion: an overview,

    M. Grandini, E. Bagli, and G. Visani, “Metrics for multi-class classifica- tion: an overview,”arXiv, Aug. 2020, doi: 10.48550/arXiv.2008.05756

    2020 doi arXiv

  39. [47]

    Comparing two K-category assignments by a K-category correlation coefficient,

    J. Gorodkin, “Comparing two K-category assignments by a K-category correlation coefficient,”Comput. Biol. Chem., vol. 28, no. 5–6, pp. 367– 374, Dec. 2004, doi: 10.1016/j.compbiolchem.2004.09.006

    2004 doi

  40. [48]

    A comparison of MCC and CEN error measures in multi-class prediction,

    G. Jurman, S. Riccadonna, and C. Furlanello, “A comparison of MCC and CEN error measures in multi-class prediction,”PLoS ONE, vol. 7, no. 8, p. e41882, Aug. 2012, doi: 10.1371/journal.pone.0041882

    2012 doi

  41. [49]

    A coefficient of agreement for nominal scales,

    J. Cohen, “A coefficient of agreement for nominal scales,”Educ. Psychol. Meas., vol. 20, no. 1, pp. 37–46, 1960, doi: 10.1177/001316446002000104. VOLUME 11, 2023 15 Rinkevičius et al.: AB-Guided Framework: Enhancing OOD Generalization in PPG-Based Sleep Apnea Characterization

    1960 doi

  42. [50]

    MS-Net: Sleep apnea detection in PPG using multi-scale block and shadow module one-dimensional convo- lutional neural network,

    K. Wei, L. Zou, G. Liu, and C. Wang, “MS-Net: Sleep apnea detection in PPG using multi-scale block and shadow module one-dimensional convo- lutional neural network,”Comput. Biol. Med., p. 106469, Jan. 2023, doi: 10.1016/j.compbiomed.2022.106469

    2023

  43. [51]

    Cuffless blood pressure measurement,

    R. Mukkamala, G. S. Stergiou, and A. P . Avolio, “Cuffless blood pressure measurement,”Annu. Rev. Biomed. Eng., vol. 24, no. 1, pp. 203–230, Jun. 2022, doi: 10.1146/annurev-bioeng-110220-014644

    2022 doi

  44. [52]

    Sleep apnea detection using pulse photoplethysmog- raphy,

    M. Deviaene, J. Lázaro, D. Huysmans, D. Testelmans, B. Buyse, S. V an Huffel, and C. V aron, “Sleep apnea detection using pulse photoplethysmog- raphy,” inProc. Comput. Cardiol. (CinC), Maastricht, Netherlands, 2018, pp. 1–4, doi: 10.22489/CinC.2018.134

    2018 doi

  45. [53]

    Generalizable deep learning for photoplethysmography-based blood pressure estimation–A benchmarking study,

    M. Moulaeifard, P . H. Charlton, and N. Strodthoff, “Generalizable deep learning for photoplethysmography-based blood pressure estimation–A benchmarking study,”Mach. Learn. Health, vol. 1, no. 1, p. 010501, Sep. 2025, doi: 10.1088/3049-477X/ae01a8. MANTAS RINKEVIČIUSreceived t...

    2016 doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.