CoChat
Sign Up
Cited by 88ECG Monitoring and AnalysisNon-Invasive Vital Sign MonitoringArtificial Intelligence in Healthcare

Artificial intelligence for detection of cardiovascular-related diseases from wearable devices: A systematic review and meta-analysis

Open access
Lee, Solam, Chu, Yuseong, Ryu, Jiseung, Park, Young Jun, Yang, Sejung, Koh, Sang Baek·2022·Yonsei Medical Journal
DOI 10.3349/ymj.2022.63.s93PMC PMC8790582
Open this paper in CoChat — free account

Abstract

PURPOSE: Several artificial intelligence (AI) models for the detection and prediction of cardiovascular-related diseases, including arrhythmias, diabetes, and sleep apnea, have been reported. This systematic review and meta-analysis aimed to identify AI models developed for or applicable to wearable and mobile devices for diverse cardiovascular-related diseases. MATERIALS AND METHODS: The searched databases included Medline, Embase, and Cochrane Library. For AI models for atrial fibrillation (AF) detection, a meta-analysis of diagnostic accuracy was performed to summarize sensitivity and specificity. RESULTS: A total of 102 studies were included in the qualitative review. There were AI models for the detection of arrythmia (n=62), followed by sleep apnea (n=11), peripheral vascular diseases (n=6), diabetes mellitus (n=5), hyper/hypotension (n=5), valvular heart disease (n=4), heart failure (n=3), myocardial infarction and cardiac arrest (n=2), and others (n=4). For quantitative analysis of 26 studies reporting AI models for AF detection, meta-analyzed sensitivity was 94.80% and specificity was 96.96%. Deep neural networks showed superior performance [meta-analyzed area under receiver operating characteristics curve (AUROC) of 0.981] compared to conventional machine learning algorithms (meta-analyzed AUROC of 0.961). However, AI models tested with proprietary dataset (meta-analyzed AUROC of 0.972) or data acquired from wearable devices (meta-analyzed AUROC of 0.977) showed inferior performance than those with public dataset (meta-analyzed AUROC of 0.986) or data from in-hospital devices (meta-analyzed AUROC of 0.983). CONCLUSION: This review found that AI models for diverse cardiovascular-related diseases are being developed, and that they are gradually developing into a form that is suitable for wearable and mobile devices.

AI snapshot

Written from the paper’s full text.

TL;DR

This systematic review and meta-analysis identified 102 studies of AI models for cardiovascular-related diseases applicable to wearable or mobile devices, with arrhythmia/AF detection dominating the literature. In 26 AF-detection studies, meta-analyzed sensitivity was 94.80% and specificity was 96.96%, though performance varied by algorithm type, data availability, and data source.

Key findings

  • Of 102 qualitatively reviewed studies, arrhythmia was the most common target (n=62), followed by sleep apnea (n=11), peripheral vascular diseases (n=6), diabetes mellitus (n=5), hyper/hypotension (n=5), valvular heart disease (n=4), heart failure (n=3), myocardial infarction and cardiac arrest (n=2), and others (n=4).
  • For 26 studies reporting AI models for AF detection, meta-analyzed sensitivity was 94.80% (95% CI, 91.94%–96.68%) and specificity was 96.96% (95% CI, 94.99%–98.17%).
  • Deep neural networks showed superior performance (meta-analyzed AUROC 0.981) compared with conventional machine learning algorithms (meta-analyzed AUROC 0.961). Models tested on public datasets had higher AUROC (0.986) than those tested on proprietary datasets (0.972), and models tested on in-hospital device data had higher AUROC (0.983) than those tested on wearable device data (0.977).
  • In the included studies, 68 used deep learning methods, 59 used an open dataset, and 47 utilized data collected through wearable devices.

Methods

  • This was a systematic review and meta-analysis performed according to PRISMA 2020, searching Medline, Embase, and Cochrane Library on September 5, 2021, without publication date limitation. Eligible studies involved AI models for detection or prediction of cardiovascular-related disease using deep learning or conventional machine learning, developed for smartphones or using data from mobile/wearable devices or relatively easy-to-obtain biosignals such as 1-lead or 2-lead ECG and PPG.
  • Two main reviewers screened titles and abstracts, with disagreements resolved through discussion with two additional reviewers. Extracted variables included publication year, author, target disease, task, data availability, data source, input data domain, algorithm type, and model performance metrics.
  • Quantitative meta-analysis was limited to AF studies. Random-effects models were used to synthesize sensitivity and specificity, and hierarchical summary ROC curves and 95% CIs were estimated using the Reitsma bivariate model. Subgroup analyses were performed by algorithm type, data availability, and data source.

Limitations

  • The paper states high heterogeneity among included studies, for example some used ICU ECG data while others used smartwatch data; although subgroup analysis was performed, inappropriate statistical estimates may have been produced by quantitatively synthesizing studies using different modalities and population groups.
  • The paper states that for diseases and tasks other than AF, meta-analysis could not be performed because there were too few studies to synthesize quantitatively.
  • Evident from the text: the QUADAS-2 tool’s validity for AI studies had not been verified and was not used to include or exclude studies; studies in languages other than English or Korean were excluded; and the literature search was performed by one author.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.3349/ymj.2022.63.s93

Cites

129 of 129 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Claims the paper makes

0 of 23 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    For quantitative analysis of 26 studies reporting AI models for AF detection, meta-analyzed specificity was 96.96%.

    “For quantitative analysis of 26 studies reporting AI models for AF detection, meta-analyzed sensitivity was 94.80% and specificity was 96.96%.”
    Quote in the paperSubstantial claim
  • Findingown data

    A total of 102 studies were included in the qualitative review.

    “A total of 102 studies were included in the qualitative review.”
    • Also stated in: § Study selection · plain
      “After a full text review, 102 studies in total were included in the qualitative review.”
    Quote in the paperSubstantial claim
  • Findingown data

    For quantitative analysis of 26 studies reporting AI models for AF detection, meta-analyzed sensitivity was 94.80%.

    “For quantitative analysis of 26 studies reporting AI models for AF detection, meta-analyzed sensitivity was 94.80% and specificity was 96.96%.”
    Quote in the paperSubstantial claim
  • Findingown data

    AI models tested with data acquired from wearable devices (meta-analyzed AUROC of 0.977) showed inferior performance than those with data from in-hospital devices (meta-analyzed AUROC of 0.983) for AF detection.

    “However, AI models tested with proprietary dataset (meta-analyzed AUROC of 0.972) or data acquired from wearable devices (meta-analyzed AUROC of 0.977) showed inferior performance than those with public dataset (meta-analyzed AUROC of 0.986) or data from in-hospital devices (meta-analyzed AUROC of 0.983).”
    Quote in the paperSubstantial claim
  • Findingown data

    AI models tested with proprietary dataset (meta-analyzed AUROC of 0.972) showed inferior performance than those with public dataset (meta-analyzed AUROC of 0.986) for AF detection.

    “However, AI models tested with proprietary dataset (meta-analyzed AUROC of 0.972) or data acquired from wearable devices (meta-analyzed AUROC of 0.977) showed inferior performance than those with public dataset (meta-analyzed AUROC of 0.986) or data from in-hospital devices (meta-analyzed AUROC of 0.983).”
    Quote in the paperNot a substantial claim
  • May combine two claims. 20
  • Classified as probably not a substantial claim. 36

  • May combine two claims.

    Findingown data

    Among included studies, the numbers of AI models for detection were: arrhythmia 62, sleep apnea 11, peripheral vascular diseases 6, diabetes mellitus 5, hyper/hypotension 5, valvular heart disease 4, heart failure 3, myocardial infarction and cardiac arrest 2, and others 4.

    “There were AI models for the detection of arrythmia (n=62), followed by sleep apnea (n=11), peripheral vascular diseases (n=6), diabetes mellitus (n=5), hyper/hypotension (n=5), valvular heart disease (n=4), heart failure (n=3), myocardial infarction and cardiac arrest (n=2), and others (n=4).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    The most common study following arrhythmia was the detection of sleep apnea.

    “The most common study following arrhythmia was the detection of sleep apnea.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Obstructive sleep apnea is associated with or increases the risk of major cardiovascular diseases.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    Owing to a large database such as the MIT-BIH arrhythmia database, various detection models for arrhythmias, including AF, have been reported.

    “Owing to a large database, such as the MIT-BIH arrhythmia database,122123 various detection models for arrhythmias, including AF, have been reported.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [122] Moody, 2001, IEEE Eng Med Biol Mag source

      Not checked in the free run.

    • cites [123] Moody, 1983, Comput Cardiol

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    A total of 59 studies used an open dataset, most of which were related to the detection of arrhythmias or sleep apnea.

    “A total of 59 studies used an open dataset, most of which were related to the detection of arrhythmias or sleep apnea.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingown data

    1-lead and 2-lead ECG were the most commonly used biosignals, followed by PPG.

    “For input domain, 1-lead and 2-lead ECG were the most commonly used biosignals, followed by PPG.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    In addition to ECG, there have been studies that use PPG, many of which used proprietary datasets.

    “In addition to ECG, there have also been studies that use PPG, many of which were studies that used proprietary datasets.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    External validation was rarely performed for diseases other than arrhythmia, considering the low availability of public data and the characteristics of studies with individual datasets.

    “Considering the low availability of public data and the characteristics of studies with individual datasets, external validation was rarely performed.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    The popularization of deep neural networks has enabled raw signals themselves to be used as input data.

    “However, the popularization of deep neural networks has enabled raw signals themselves to be used as input data.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Findingnone given

    Several studies used raw ECG with deep learning algorithms, but machine learning models using handcrafted features such as heart rate variability have also been reported.

    “There have been several studies that used raw ECG with deep learning algorithms, but machine learning models using handcrafted features, such as heart rate variability, have also been reported.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    The public dataset PhysioNet Apnea-ECG has been used in numerous studies.

    “The public dataset PhysioNet Apnea-ECG116 has been used in numerous studies.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [116] Penzel, 2000, Computers in Cardiology

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    A model that could detect AF in a large number of patients using facial video rather than ECG monitoring was reported, confirming that diverse types of input data could be utilized.

    “In addition, a model that could detect AF in a large number of patients using a video of the facial region rather than ECG monitoring was reported, confirming that diverse types of input data could be utilized.36”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [36] Yan, 2020, JAMA Cardiol source

      Not checked in the free run.

  • May combine two claims.

    Findingcitation

    Studies using ring-type or patch-type devices showed satisfactory performance in general but slightly lower performance for detecting arrhythmias compared to models trained with resting ECG-based data obtained in the hospital.

    “These studies have also shown satisfactory performance in general, but reported slightly lower performance for detecting arrhythmias compared to the models trained with resting ECG-based data obtained in the hospital.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    Although many studies utilized public datasets such as MIT-BIH, some studies performed external validation by constructing a proprietary dataset with multi-institutional data.

    “Although many studies have utilized public datasets, such as the MIT-BIH database alone, there were also studies that performed external validation by constructing a proprietary dataset with multi-institutional data.17232627”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    Biosignals were often used as a single input source, but some studies used multimodal data.

    “They were often used as a single input source, but some studies used multimodal data.232898”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingcitation

    A study using wrist-worn reflective PPG reported that false positive detection had an adverse effect on the specificity and positive predictive value of the model.

    “This particular study investigated the changes in model performance due to limbic movement, and reported that false positive detection had an adverse effect on the specificity and positive predictive value of the model.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [79] Papini, 2020, Sci Rep source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    Various machine learning methods have been used to predict cardiovascular disease using handcrafted features such as heart rate variability, R-R interval, and QRS amplitude extracted from ECG or PPG.

    “Various machine learning methods have been used to predict cardiovascular disease using handmade features such as heart rate variability, R-R interval, and QRS amplitude extracted from ECG or PPG as general input values.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingreasoning

    A large number of models have been reported based on the rich public dataset, which allowed performance comparisons between the studies.

    “A large number of models have been reported based on the rich public dataset, which allowed performance comparisons between the studies.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based). May combine two claims.

    Findingnone given

    The most commonly used biosignal for monitoring using a wearable device is the 1-lead or 2-lead ECG signal, followed by PPG.

    “The most commonly used biosignal for monitoring using a wearable device is the 1-lead or 2-lead ECG signal, followed by PPG.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingnone given

    There have been a considerable number of models that used CNN alone or CNN combined with RNN-based algorithms.

    “However, there have been a considerable number of models that used convolutional neural network (CNN) alone or CNN combined with RNN-based algorithms.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Findingcitation

    A small number of studies have adopted salient maps to improve interpretability, but such attempts are still difficult to find compared to medical imaging models.

    “While a small number of studies have adopted a strategy that uses salient maps to improve interpretability,98113 such attempts are still difficult to find compared to the models in the field of medical imaging.”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [98] Lee, 2021, Br J Anaesth source

      Not checked in the free run.

    • cites [113] Kwon, 2020, Lancet Digit Health source

      Not checked in the free run.

  • May combine two claims.

    Findingown data

    Some included studies used ECG data from an intensive care unit, while others used data obtained with a smart watch.

    “For instance, some studies used ECG data obtained from an intensive care unit, whereas some studies directly utilized data obtained with a smart watch.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Findingnone given

    Numerous studies have demonstrated that deep learning algorithms show very high performance compared to existing analysis methods that use human visualization or extraction of handcrafted features for biosignals such as ECG or PPG.

    “Numerous studies have demonstrated that the deep learning algorithm shows very high performance compared to the existing analysis methods that use human visualization or the extraction of handmade features for biosignals, such as ECG or PPG signals.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationown data

    AI models for diverse cardiovascular-related diseases are being developed and are gradually developing into a form suitable for wearable and mobile devices.

    “This review found that AI models for diverse cardiovascular-related diseases are being developed, and that they are gradually developing into a form that is suitable for wearable and mobile devices.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    It is difficult to cope with the burden and death of cardiovascular disease by treatment in hospital alone.

    “Therefore, it is difficult to cope with the burden and death of cardiovascular disease by treatment in hospital alone.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Most cardiovascular diseases follow a chronic course.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    In case of disagreement between the two main reviewers, eligibility was decided through discussion with two additional reviewers.

    “In the case of disagreement between the two main reviewers, eligibility was finally decided through discussion with two additional reviewers (J.S.R and S.Y).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Cardiovascular disease is the most common cause of death in the United States.

    “Cardiovascular disease is the most common cause of death in the United States,4”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [4] Centers for Disease Control and Prevention. Underlying Cause of Death, 1999-2019. [Internet]. accessed on 2021 October 1. Available at: https://wonder.cdc.gov/ucd-icd10.html

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductcitation

    This systematic review and meta-analysis was performed in accordance with PRISMA 2020 guidelines.

    “This systematic review and meta-analysis was performed in accordance with the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020.7”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [7] Page, 2021, BMJ source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Premisecitation

    A systematic literature review revealed that deep learning models for analyzing ECG showed better performance compared to existing methods, such as hardware or rule-based algorithms, for detection or prediction of cardiovascular disease.

    “A systematic literature review revealed that deep learning models for analyzing ECG showed better performance compared to existing methods, such as the hardware or rule-based algorithms used for the detection or prediction of cardiovascular disease.6”
    Quote in the paperSubstantial claim0 of 1 sources checked
    • cites [6] Krittanawong, 2019, Eur Heart J source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Cardiovascular disease incurs annual medical costs of up to $363 billion.

    “and it incurs annual medical costs of up to $363 billion.5”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [5] Virani, 2021, Circulation source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The abstracts and titles of retrieved studies were screened by two main reviewers.

    “The abstracts and titles of the retrieved studies were first screened by two main reviewers (S.L and Y.C).”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    Because biosignal data has a time-series characteristic and is not limited in input signal length, RNN or LSTM is logically the most preferred model.

    “Since biosignal data has a time-series characteristic and is not limited in the length of an input signal, recurrent neural network (RNN) or long short-term memory is logically the most preferred model.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    A random-effects model was used for synthesis because high heterogeneity between studies was suspected.

    “As high heterogeneity between studies was suspected, a random-effects model was used for synthesis.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationcitation

    With the advent of deep neural networks, competition for predictive performance has virtually reached saturation, and development of lightweight models for wearable devices has become a major research topic.

    “With the advent of deep neural networks, the competition for predictive performance has virtually reached saturation, and the development of a lightweight model that can be mounted on a wearable device and operated in real time has recently become a major topic of research.1827”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationreasoning

    The availability of ambulatory biosignals is expected to increase significantly in the future.

    “Accordingly, the availability of ambulatory biosignals is expected to increase significantly in the future.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    This systematic review and meta-analysis summarized AI models developed for or potentially useful for wearable devices for detection or prediction of cardiovascular-related diseases.

    “This systematic review and meta-analysis summarized the AI models that were developed for or that are potentially useful for wearable devices, and that could be used to detect or predict cardiovascular-related diseases.”
    Quote in the paperSubstantial claim

  • May combine two claims.

    Premisecitation

    The prevalence of cardiovascular diseases varies greatly depending on the population, and class imbalance between normal and disease data may have a significant impact on model performance.

    “For example, the prevalence of cardiovascular diseases varies greatly depending on the population, and the class imbalance between normal data and disease data may have a significant impact on the model performance.108”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [108] Wang, 2019, Sensors (Basel) source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductreasoning

    Subgroup analysis was performed according to algorithm type, data availability, and data source to resolve heterogeneity among studies in the meta-analysis.

    “In order to resolve heterogeneity among studies in the meta-analysis, subgroup analysis was performed according to the algorithm type (conventional machine learning vs. deep neural network), data availability (public vs. proprietary), and data source (in-hospital device vs. wearable device).”
    Quote in the paperSubstantial claim

  • Classified as probably not a substantial claim.

    Interpretationnone given

    It is difficult to operate a large-scale model on portable devices due to hardware limitations.

    “It is difficult to operate a large-scale model on portable devices due to hardware limitations.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: It is difficult to operate a large-scale model on portable devices due to hardware limitations.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    If too many false positive alarms for abnormal signals occur, the user may prefer to stop using the wearable device.

    “If too many false positive alarms for these abnormal signals occur, the user may prefer to stop using the wearable device.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: If too many false positive alarms for abnormal signals occur, the user may prefer to stop using the wearable device.

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationcitation

    Wearable device-enabled detection of cardiovascular-related diseases is likely to become more common as healthcare technology expands.

    “Wearable device-enabled detection of cardiovascular-related diseases are likely to become more common as healthcare technology expands.139”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [139] Bayoumy, 2021, Nat Rev Cardiol source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing), S4 (based).

    Interpretationnone given

    The most significant problem with deep learning is its black box-like characteristic that makes understanding its operation and judgment principle difficult.

    “Nevertheless, the most significant problem with deep learning is that it has a black box-like characteristic that makes understanding its operation and judgment principle difficult.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Deep learning has a black box-like characteristic that makes understanding its operation and judgment principle difficult.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    It is estimated that more than 50 million people in the United States use wearable devices to record their daily activities.

    “It is estimated that more than 50 million people in the United States use wearable device to record their daily activities.140”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [140] Steinberg, 2017, Heart Rhythm source

      Not checked in the free run.

  • Classified as probably not a substantial claim. May combine two claims.

    Interpretationreasoning

    Lifelogs from wearable devices could lead to early detection of diverse cardiovascular-related diseases and potentially life-threatening conditions.

    “These lifelogs could lead to the early detection of diverse cardiovascular-related diseases and potentially life-threatening conditions.”
    Quote in the paperSubstantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S4 (based).

    Interpretationnone given

    Unlike medical imaging, AI models for cardiovascular-related disease are preferred to be used with wearable and portable devices carried by individuals in their day-to-day lives.

    “Unlike medical imaging, AI models for cardiovascular-related disease are preferred to be used with wearable and portable devices carried by individuals in their day-to-day lives.”
    Quote in the paperNot a substantial claim

  • Classified as probably not a substantial claim.

    Interpretationreasoning

    The algorithm is operated in an environment that is not well-controlled, unlike a hospital environment.

    “Second, the algorithm is operated in an environment that is not well-controlled, unlike a hospital environment.”
    Quote in the paperSubstantial claim
    • Rests on an assumption: The algorithm is operated in an environment that is not well-controlled, unlike a hospital environment.

  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Interpretationreasoning

    Although patient-activated daily monitoring has the potential benefit to improve detection of subclinical or occult diseases, there are limitations of their widespread use.

    “However, although patient-activated daily monitoring has the potential benefit to improve the detection of subclinical or occult diseases, there are limitations of their widespread use.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S2 (checkable), S3 (load-bearing).

    Interpretationreasoning

    For the analysis of lifelogs collected in real time, the operating speed is also a major consideration.

    “For the analysis of lifelogs collected in real time, the operating speed is also a major consideration.”
    Quote in the paperNot a substantial claim

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Interpretationnone given

    Even resting ECG signals may be more prone to artifacts or noise if they are acquired in daily life.

    “In addition, even resting ECG signals may be more prone to artifacts or noise if they are acquired in daily life.”
    Quote in the paperNot a substantial claim
    • Rests on an assumption: Even resting ECG signals may be more prone to artifacts or noise if they are acquired in daily life.

  • Probably not a substantial claim: may not pass S2 (checkable).

    Interpretationreasoning

    Despite the high accuracy of deep learning models, the decisions cannot be accepted by healthcare workers without proper interpretation.

    “Despite the high accuracy of deep learning models, the decisions cannot be accepted by healthcare workers without proper interpretation.”
    Quote in the paperNot a substantial claim

  • May combine two claims.

    Interpretationcitation

    Poor signal quality and false alarms can lead to inappropriate interpretation, resulting in unnecessary medical referrals and testing.

    “For example, poor signal quality and false alarms can lead to inappropriate interpretation, resulting in unnecessary medical referrals and testing.141”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [141] Bumgarner, 2018, J Am Coll Cardiol source

      Not checked in the free run.

50 of 50

  • The paper relies on this without support.

    Most cardiovascular diseases follow a chronic course.

    “Since most cardiovascular diseases follow a chronic course, lifestyle modifications and the periodic monitoring of diseases are very important even after acute treatment in hospital.”

  • The paper relies on this without support.

    Obstructive sleep apnea is associated with or increases the risk of major cardiovascular diseases.

    “Sleep apnea itself is not a cardiovascular disease, but obstructive sleep apnea is a major detection target for prediction models for cardiovascular-related disease that use wearable devices, as it is associated with or increases the risk of major cardiovascular diseases.”

  • The paper relies on this without support.

    There has been no optimal assessment tool that can evaluate the applicability and risk of bias for AI studies.

    “So far, there has been no optimal assessment tool that can evaluate the applicability and risk of bias for AI studies.”

  • The paper relies on this without support.

    The validity of using the QUADAS-2 scale has not been verified for AI studies.

    “However, since the validity of using the scale has not been verified for AI studies, it was not used to determine whether to include or exclude a particular study.”

  • The paper relies on this without support.

    Deep learning has a black box-like characteristic that makes understanding its operation and judgment principle difficult.

    “Nevertheless, the most significant problem with deep learning is that it has a black box-like characteristic that makes understanding its operation and judgment principle difficult.”

  • The paper relies on this without support.

    It is difficult to operate a large-scale model on portable devices due to hardware limitations.

    “It is difficult to operate a large-scale model on portable devices due to hardware limitations.”

  • The paper relies on this without support.

    The algorithm is operated in an environment that is not well-controlled, unlike a hospital environment.

    “Second, the algorithm is operated in an environment that is not well-controlled, unlike a hospital environment.”

  • The paper relies on this without support.

    Even resting ECG signals may be more prone to artifacts or noise if they are acquired in daily life.

    “In addition, even resting ECG signals may be more prone to artifacts or noise if they are acquired in daily life.”

  • The paper relies on this without support.

    If too many false positive alarms for abnormal signals occur, the user may prefer to stop using the wearable device.

    “If too many false positive alarms for these abnormal signals occur, the user may prefer to stop using the wearable device.”
9 of 9

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • There are no systematic reviews addressing the potential of smart wearables for early diagnosis of CVDs.

    “To the best of our knowledge, there are no systematic reviews addressing the potential of smart wearables for early diagnosis of CVDs. For example, in [34], the authors investigated the application of AI in smart wearables for cardiovascular disease detection. However, the focus of their research was on smart models rather than hardware; the obstacles that have slowed the development of this field”

    From Smart Wearables for the Detection of Cardiovascular Diseases: A Systematic Literature Review (2023) · cites this paper as [34]

Cites

Flagged first, then by citations

Research works printed but not linked: 30

  1. Horjus, 2011, Cochrane Database Syst Rev, 2011, CD005184
  2. Taylor, 2013, Cochrane Database Syst Rev, 2013, CD004816
  3. Shi, 2021, Comput Math Methods Med, 2021, 6691177
  4. Lu, 2021, J Healthc Eng, 2021, 6684954
  5. Tan, 2021, Neural Comput Appl
  6. Gao, 2020, Annu Int Conf IEEE Eng Med Biol Soc, 2020, 316
  7. Vijayarangan, 2020, Annu Int Conf IEEE Eng Med Biol Soc, 2020, 300
  8. Cheng, 2020, Annu Int Conf IEEE Eng Med Biol Soc, 2020, 580
  9. Lennox, 2020, Annu Int Conf IEEE Eng Med Biol Soc, 2020, 288
  10. Abdelazez, 2020, Annu Int Conf IEEE Eng Med Biol Soc, 2020, 5398
  11. Kwon, 2019, Arrhythmia, 35, 104 AP19-00184
  12. Wu, 2019, Annu Int Conf IEEE Eng Med Biol Soc, 2019, 1908
  13. Kwon, 2019, Heart Rhythm, 16, S578 S-IA01-013
  14. Lai, 2019, Annu Int Conf IEEE Eng Med Biol Soc, 2019, 4897
  15. Hernandez, 2018, Annu Int Conf IEEE Eng Med Biol Soc, 2018, 5982
  16. Khamis, 2018, Annu Int Conf IEEE Eng Med Biol Soc, 2018, 5998
  17. Zhou, 2018, Annu Int Conf IEEE Eng Med Biol Soc, 2018, 2551
  18. Yan, 2018, Heart Rhythm, 15, S362 B-PO03-185
  19. Clifford, 2017, Comput Cardiol (2010)
  20. Jin, 2009, Annu Int Conf IEEE Eng Med Biol Soc, 2009, 6889
  21. Iwasaki, 2021, Sleep Breath
  22. Wang, 2019, Biomed Res Int, 2019, 9768072
  23. Rashid, 2020, Annu Int Conf IEEE Eng Med Biol Soc, 2020, 4648
  24. Penzel, 2000, Computers in Cardiology, 27, 255
  25. Goldberger, 2000, Circulation, 101, E215
  26. Moody, 1983, Comput Cardiol, 10, 227
  27. Moody, 1984, Comput Cardiol, 11, 381
  28. Iyengar, 1996, Am J Physiol, 271, R1078
  29. Moody, 2001, Computers in Cardiology, 28, 113
  30. Bousseljot, 1995, Biomed Eng, 40, 317

Cited by

Sorted by citations

Related

Authors
Lee, S., Chu, Y., Ryu, J., Park, Y. J., Yang, S., & Koh, S. B.
Affiliations
Yonsei University (KR)
Corresponding author
  • Lee, Solam
  • Chu, Yuseong
ORCID
Venue
Yonsei Medical Journal
Publisher
XMLink
Published
2022
Volume
63
Issue
Suppl
Pages
S93
ISSN
  • 0513-5796 (ISSN-L)
  • 1976-2437
DOI
10.3349/ymj.2022.63.s93
PMC
PMC8790582
OpenAlex
W4205432209
Semantic Scholar
e0837c30ec39becb2710974a76afbb606b8b3454
Cited by
88
References
108
Citations by year
2026: 13 · 2025: 28 · 2024: 23 · 2023: 16 · 2022: 8
Type
conference abstract
Language
en
Open access
Licence
Locations
  • Yonsei Medical Journal (publishedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by-nc)
  • PubMed (publishedVersion)
Funders
Korea Health Industry Development Institute
Topics
  • Health Sciences › Medicine › Cardiology and Cardiovascular Medicine › ECG Monitoring and Analysis
  • Physical Sciences › Engineering › Biomedical Engineering › Non-Invasive Vital Sign Monitoring
  • Health Sciences › Health Professions › Health Information Management › Artificial Intelligence in Healthcare
Keywords
artificial intelligence, cardiovascular disease detection, wearable devices, arrhythmia detection, atrial fibrillation, sleep apnea, diabetes mellitus, deep neural network, diagnostic accuracy, heart failure, peripheral vascular disease, hypertension, valvular heart disease, mobile health
MeSH terms
Wearable Electronic Devices, Algorithms, Artificial Intelligence, Atrial Fibrillation, Humans, Sensitivity and Specificity

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-09.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-09.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review