Digital cardiology: A state-of-the-art review of wearable device applications for diagnosis, monitoring, and management of cardiovascular diseases

Abstract

Background

Wearable devices are increasingly used in cardiovascular medicine, but their clinical applications, evidentiary basis, and implementation challenges require comprehensive assessment.

Methods

We conducted a narrative review by searching PubMed, EMBASE, and the Cochrane Library for articles published since 2020. Potentially relevant papers were screened and compared for relevance, followed by independent assessment for inclusion. The focus is on devices with direct clinical relevance and available human studies.

Results and Conclusions

Wearable devices show promise for arrhythmia detection (particularly atrial fibrillation), blood pressure monitoring, heart failure management, and cardiac rehabilitation. Commercial devices like smartwatches have demonstrated utility in large-scale studies, while emerging technologies like smart textiles and implantable sensors offer future potential. Key challenges include validation against gold-standard methods, integration into clinical workflows, data privacy concerns, and ensuring health equity. For successful implementation, future efforts should focus on robust clinical trials, regulatory clarity, interoperability standards, and clinician education. Wearable technologies are poised to transform cardiovascular care through continuous, personalized monitoring, but require multidisciplinary collaboration to realize their full clinical potential.

1. Introduction

Cardiovascular disease (CVD) is the leading cause of death and disability worldwide, with a lifetime risk exceeding 60%. ¹ Ischemic heart disease and stroke are the primary contributors to disability-adjusted life-years (DALYs) in adults aged 50 and above. ² In the United States, nearly half of all adults aged 20 or older have some form of CVD—including coronary heart disease (CHD), heart failure (HF), stroke, or hypertension—making it a greater cause of mortality than cancer in those under 85¹. Similarly, in China, CVD is the top killer in both rural and urban regions, responsible for approximately 48% of all deaths, and its prevalence continues to rise.^3,4

CVD remains a major threat to global health, underscoring the need for effective prevention and treatment. The advent of wearable technology has opened new possibilities for monitoring and preventing cardiovascular diseases.

Since their inception in the 1960s by Edward O'Thorp, ⁵ wearable devices have evolved from academic concepts into versatile tools used across medical, educational, and military fields. ⁶ In healthcare, devices such as smart bracelets and activity trackers now play a significant role in cardiovascular diagnosis and monitoring. These tools capture physiological data through mobile sensors and apply specialized algorithms to support disease prevention and management. ⁷ Some advanced systems also enable localized drug delivery, paving the way for intelligent treatment. ⁸

However, significant challenges and limitations in wearable technology hinder its broader medical application and impede the advancement of intelligent digital diagnostics. Persistent gaps remain in continuous monitoring for many diseases. The current literature lacks technical reviews tailored for clinicians, clear distinctions regarding levels of evidence, and a perspective on clinical integration. Therefore, this review aims to: (1) summarize current clinical applications of wearable devices in cardiovascular medicine; (2) evaluate the level of clinical validation for each technology; and (3) explores emerging cutting-edge technologies with clinical potential (Tables 1 and 2).

Table 1.

Classification and applications of wearable devices in cardiovascular medicine.

Category	Device/Technology	Primary function(s)	Key advantages	Current limitations
Monitoring & Diagnosis	Smartwatches/Fitness Trackers (PPG, 1L-ECG)	HR, HRV, AF detection, Activity tracking	Convenience, continuous monitoring, user engagement	Motion artifacts, lower accuracy vs. clinical ECG, skin tone bias
	Cuffless BP Monitors (PPG/PAT-based)	Continuous BP estimation	Non-invasive, ambulatory monitoring	Requires calibration, accuracy affected by activity/posture
	Smart Socks/Insoles (Temperature/Pressure)	Diabetic foot ulcer prediction	Early warning, home-based care	High false positive rate, durability issues, single-leg use limitation
	IMU-based Gait Analyzers	Posture, balance, gait analysis	Portable, real-time feedback, long-term tracking	Limited prognostic evidence, need for standardized metrics
Treatment & Intervention	Bioimpedance Smart Dressings	Wound healing assessment, infection monitoring	Objective, continuous monitoring under dressing	Early stage, needs clinical trial validation
	Microneedle Array (MNA) Smart Bandages	Targeted drug/growth factor delivery	Programmable, minimal invasiveness, promotes healing	High cost, biocompatibility challenges, only animal studies so far
Clinical Support & Training	AR/VR Surgical Navigation	Anatomical overlay, procedural guidance	Enhanced spatial orientation, reduced errors	High cost, integration into workflow, data privacy
	VR Surgical Simulators	Immersive surgical training	Risk-free practice, skill acquisition	Lack of standardized evaluation, ethical concerns

Table 2.

Clinical technologies for wearable devices in cardiovascular medicine.

Device type	Technology	Modality	Clinical availability	Evidence level*	Clinical outcome data available
Smartwatch	Blood pressure & AF detection	PPG	Yes	2b	Yes
	AF detection (ongoing trial)	PPG	Yes	N/A (ongoing)	No (pending)
	Heart failure exacerbation prediction	Multisensor patch	Investigational	2b	Yes
	AF & arrhythmia screening	PPG	Yes	2b	Yes
	Heart rate control evaluation	PPG	Yes	3b	Yes
	Energy-efficient HR monitoring	PPG + machine learning	Investigational	3a	No
	Cuffless BP (pulse arrival time)	PAT-based wearable	Investigational	3b	Yes
	Motion artifact reduction	PPG + interface sensor	Investigational	N/A	No
	Cuffless BP challenges	PPG + ECG	Yes	3b	Yes
	Cuffless BP (pattern fusion)	PPG + ECG + fusion model	Investigational	3b	Yes
	HRV analysis	HRV + error compensation	Investigational	3b	Yes
	Vascular age/stiffness assessment	PPG	Investigational	3a	Yes
	Aortic stiffness estimation	PPG + ECG + machine learning	Investigational	3b	Yes
Patch/Smartwatch	Ultrasound-based monitoring	CMUT/PMUT	Investigational	3a	No
Smart bandage	microneedle drug delivery	Microneedle array	Investigational	N/A	No
	AI wound monitoring	Neural network + RF tuning	Investigational	N/A	No
	closed-loop	Bioimpedance + temperature + electrical stimulation	Investigational	N/A	No
Smart camera	Contact-free AF detection	Camera-based FPPG + deep learning	Investigational	3b	Yes
Temperature sensor mat	Diabetic foot ulcer prediction (mat)	Temperature sensors	Investigational	4	Yes
Balance and gait monitoring equipment	Gait & balance (neurological/diabetic)	IMU	Investigational	3a	Yes
	Gait in diabetic foot management	IMU	Investigational	3b	Yes
	Real-time gait detection	IMU	Investigational	3a	Yes
fitness trackers/activity-tracking smartwatches/and pedometers	Physical activity tracking	Inertial sensors	Yes	2a	Yes
Smart socks	Diabetic foot ulcer prediction	Temperature sensors	Yes	3b	No
Smart insoles	Diabetic foot ulcer prediction	Pressure + temperature sensors	Investigational	N/A	No
Smart dressing	Wound moisture monitoring	Humidity sensor	Investigational	3b	Yes
	Venous ulcer monitoring	Bioimpedance	Investigational	4	Yes
Wearable sphygmomanometer	Ambulatory BP monitoring	Pressure sensor (e.g., Omron HeartGuide)	Yes	2a	Yes
	BP waveform estimation	Pressure sensor/ultrasound/ML	Investigational	3a	Yes
Virtual surgery	Surgical telementoring	AR (e.g., HoloLens)	Investigational	N/A	No
Smart glasses/headgear	VR/AR in clinical care	VR, AR	Investigational	3a	No
Wearable system	Multimodal wearable systems	Multi-sensor (lower extremity vascular)	Yes	2a	Yes
Wearable material	Flexible materials for wearables	Flexible electronics	Investigational	3a	No

PPG = photoplethysmography; ECG = electrocardiogram; HRV = heart rate variability; IMU = inertial measurement unit; AR = augmented reality; VR = virtual reality; FPPG = facial photoplethysmography; PAT = pulse arrival time; ML = machine learning.

^*Evidence level based on Oxford Centre for Evidence-Based Medicine (CEBM) criteria: 2a = Systematic review of cohort studies; 2b = Individual cohort study; 3a = Systematic review of case-control studies; 3b = Individual case-control study; 4 = Case-series; N/A = Not applicable (preclinical, ongoing trial, or insufficient evidence).

2. Methods

2.1. Literature search strategy

Given that this field is still in its infancy, technology evolves rapidly, and research methodologies vary significantly, we opted for a narrative review approach. We conducted a comprehensive literature search of PubMed, EMBASE, and the Cochrane Library for articles published between January 1, 2020 and July 31, 2025. The search strategy combined controlled vocabulary (MeSH terms for PubMed; Emtree terms for EMBASE) and title/abstract keywords for wearable technologies and cardiovascular diseases.

2.2. Inclusion and exclusion criteria

Studies were included if they met the following criteria: (1) original research including randomized controlled trials, prospective or retrospective cohort studies, case-control studies, or validation studies; (2) systematic reviews or meta-analyses; (3) focus on wearable device applications in cardiovascular medicine; (4) English language; (5) human subjects.

Studies were excluded if they met any of the following criteria: (1) animal studies or in vitro studies; (2) pure engineering or algorithm development without clinical validation data; (3) conference abstracts (unless seminal); (4) editorials, commentaries, or opinion pieces; (5) studies focused solely on non-cardiovascular applications.

2.3. Study selection

The literature search initially yielded 182 records after duplicate removal. Two reviewers (QY and TZ) independently screened the titles and abstracts of all retrieved records. During this phase, 58 records were excluded because they: (1) were not related to cardiovascular medicine; (2) did not focus on wearable devices; (3) were conference abstracts, editorials, or commentaries; or (4) were animal studies.

The remaining 124 articles underwent full-text review. After thorough evaluation, an additional 81 articles were excluded due to insufficient data quality, lack of standardized methodology, or limited clinical relevance. Ultimately, 43 articles met all inclusion criteria and were included in the final synthesis. Any disagreements between the two reviewers during the screening process were resolved through discussion. The study selection process is documented in the flow diagram (Figure 1).

Figure 1.

Flow diagram of the study selection process.

2.4. Data synthesis

Given the heterogeneity of study designs and outcome measures, a narrative synthesis approach was adopted. Studies were grouped by clinical application: (1) diagnostic (atrial fibrillation detection, blood pressure monitoring); (2) monitoring (heart failure, diabetic foot ulcer prevention, gait analysis); (3) therapeutic (smart dressings, drug delivery systems); and (4) clinical support (AR/VR for surgical training and navigation). Within each group, findings were synthesized qualitatively with attention to evidence level, study quality, and clinical relevance.

2.5. Quality and bias considerations

While formal quality assessment tools (e.g., QUADAS-2, ROBINS-I) were not applied due to the narrative nature of this review, we considered the following factors during data interpretation: study design (prioritizing RCTs and prospective cohorts over retrospective or case-series), sample size (studies with <50 participants noted as preliminary), potential conflicts of interest (industry sponsorship noted where applicable), and regulatory status (FDA/CE clearance noted where available).

3. Main text

3.1. Current clinical applications of wearable devices

3.1.1. Real-time measurement of body data

3.1.1.1. Physiological data monitoring

Using sensors and digital software, wearable devices accurately track metrics such as heart rate, blood pressure, ECG, blood oxygen, sleep, and stress levels. In cardiovascular surgery, monitoring prioritizes heart rate, ECG, blood pressure, and blood oxygen saturation.

Heart rate (HR), influenced by factors such as age, genetics, and psychological state, serves as an important indicator of cardiovascular health. Most wearable devices measure HR using single-lead ECG (1L-ECG), ⁹ which captures signals similar to lead I of a standard 12L-ECG by holding two metal electrodes with two hands or multiple fingers of each hand. ¹⁰ Compared to 12L-ECG, it enables convenient, continuous monitoring—especially beneficial for elderly or homebound patients—and can effectively detect atrial fibrillation and certain arrhythmias. However, 1L-ECG offers limited utility in identifying coronary syndrome, myocardial infarction, or left ventricular hypertrophy. ¹⁰ Although Simrat K. Gill et al. suggest that wearable monitoring may replace some in-person arrhythmia assessments, ¹¹ practical challenges remain, including skin irritation from electrodes and difficulty distinguishing signal from noise during episodes such as seizures. ⁹

Additionally, some wearables employ photoplethysmography (PPG), a low-cost optical method that detects blood volume changes in tissue micro vessels. Its core mechanism uses a light source and detector positioned on opposite sides, ¹² operating in either transmission or reflection mode—the latter offering greater flexibility in placement and higher signal stability. The difference lies in whether the light emitted by the light source passes through the body to reach the light detector. ¹³ By measuring light absorption variations caused by pulsatile blood flow, PPG captures complex waveforms influenced by cardiac, respiratory, and autonomic nervous activity. While practical for rapid physiological monitoring, the raw PPG signal is inherently complex and susceptible to noise. ¹⁴

It is crucial to distinguish between the two primary sensing modalities used in wearable devices. ECG directly records the heart’s electrical activity via electrodes placed on the skin’s surface, thereby providing a direct measurement of heart rhythm and conduction. In contrast, PPG is an optical technology that detects changes in pulsatile blood volume within peripheral microvessels, thereby indirectly estimating heart rate and rhythm. Although PPG enables passive, continuous monitoring without requiring active user participation, it is more susceptible to motion artifacts, ambient light interference, and skin pigmentation compared to ECG. Therefore, PPG-based devices are best suited for screening and trend monitoring, while ECG remains indispensable when a definitive diagnosis is required. ⁹

A key premise in Takuo Aoyagi’s development of pulse oximetry was that only arterial blood modulates the PPG signal. ¹⁵ This modulated component is termed AC, while constant background absorption is DC. Due to difficulties in standardizing DC, PPG waveforms cannot be directly compared between individuals. Nevertheless, PPG reflects heart rate, blood oxygen, blood volume, and vascular status, though it is highly sensitive to interference such as motion, ambient light, tissue properties, and low perfusion. Its accuracy diminishes in obese individuals or those with vascular diseases due to poor signal penetration. ¹⁶

Heart rate variability (HRV), which measures fluctuations in RR intervals, serves as a key indicator of autonomic balance and cardiovascular risk. ¹⁷ Traditionally assessed via ECG under controlled conditions, HRV measurement is challenging in wearable devices. ECG measurement typically requires either long-term (at least 18 hours) Holter monitoring or short-term (usually five minutes) ECG recordings obtained under controlled, standardized conditions.^9,18 While PPG-based HRV shows high correlation with ECG-derived values and holds potential for autonomic monitoring and early shock detection, it remains limited by vascular and motion artifacts. Current PPG-HRV methods do not yet meet American Heart Association’s recommended measurement standards, ¹² but error compensation algorithms have shown promise in improving accuracy across time, frequency, and nonlinear domains. ¹⁹

Blood pressure (BP) monitoring is essential in cardiovascular care, yet traditional methods require stationary patients and clinical settings, limiting continuous daily use. Wearable devices now leverage PPG technology to estimate BP non-invasively and continuously by analyzing pulse wave characteristics. The approach relies on physiological principles: rising BP increases arterial stiffness and pulse wave velocity (PWV), altering peripheral PPG waveforms which correlate with central BP. ¹⁶ However, accurate BP estimation from PPG remains challenging. While machine learning (ML) can reduce measurement errors, achieving a standard deviation under 8 mmHg is difficult.^20,21 Moreover, ML models often struggle to adapt to individual physiological variations—a critical requirement for clinical usefulness.

Yali Zheng et al. improved cuff less BP estimation using pulse arrival time (PAT), significantly enhancing accuracy for nighttime systolic (error: -1.4 ± 6.6 mmHg) and diastolic (0.4 ± 6.7 mmHg) measurements. ²² PAT is the time it takes for the pressure pulse to travel from the heart to the periphery and is considered a potential substitute for blood pressure. ²³ However, daytime monitoring was less reliable due to variations in vascular tone and pre-ejection period (PEP) caused by activity, posture, and meals. Performance also declined in hypertensive patients, likely due to persistent hormonal and structural vascular changes. Key limitations include the sensitivity of PAT to confounding factors—like those seen in cold pressor tests—which may mask true BP changes. ²¹ The study’s small sample underscores the need for broader validation. Furthermore, the absence of a standardized validation method for cuff less devices remains a major translational challenge.

Shirong Qiu et al. introduced a pattern fusion (PF) method that integrates cardiovascular coupling effects through three computational modules. ²⁴ By combining BP features from ECG and PPG signals with cardiac features for correction, the PF approach achieved high blood pressure prediction accuracy across diverse scenarios and durations, outperforming single-signal methods. While algorithmic enhancements like this show promise for improving cuff less BP monitoring, challenges persist in complex real-world conditions and long-term accuracy decay. Incorporating additional data—such as age, gender, and health status—with AI-driven self-correction may support more sustainable monitoring and proactive cardiovascular prevention.

In addition to screening, wearable devices may play an emerging role in the management of atrial fibrillation. Continuous monitoring can assess the atrial fibrillation burden, which is associated with stroke risk, and guide treatment decisions, such as choosing between rate control and rhythm control strategies. Heart rate trends obtained from wearable devices can also help assess the efficacy of heart rate control medications, as demonstrated by the RATE-AF trial. ¹¹ Furthermore, long-term monitoring aids in detecting AF recurrence following catheter ablation or cardiac resynchronization therapy, potentially reducing the need for routine intermittent Holter monitoring. However, it should be noted that current guidelines do not recommend initiating anticoagulation based solely on AF detection by wearable devices, unless confirmed by ECG, and the impact of wearable-guided management on reducing hard endpoints such as stroke remains to be demonstrated by ongoing trials.

3.1.1.2. Gait and balance detection

Gait analysis, traditionally confined to costly and time-consuming lab-based systems, remains inaccessible to many patients.^25,26 Wearable devices with inertial measurement units (IMUs)—incorporating accelerometers, gyroscopes, and magnetometers—now enable portable, real-time posture and motion tracking. ²⁷ These sensors integrate accelerometers and gyroscopes within an IMU to capture linear acceleration and angular velocity, respectively. Clinically, a 3D system tracking yaw, pitch, and roll serves as the reference. Compact, energy-efficient, and portable, IMUs enable real-time posture monitoring and are particularly suited for extended gait assessment, as they minimize artifacts from gait initiation and termination. This supports continuous disease progression tracking in practical settings. ²⁷ Gait monitoring using IMUs holds particularly significant clinical value for two groups of patients with vascular diseases. First, for diabetic patients at risk of foot ulcers, subtle changes in gait often precede ulcer formation and may even predict its occurrence, thereby providing a basis for early behavioral intervention. Second, for patients with peripheral artery disease and intermittent claudication, wearable IMUs can continuously monitor walking distance and gait symmetry during daily activities, capturing claudication patterns that are difficult to assess in a laboratory setting, and providing support for personalized rehabilitation. ²⁸ However, as an emerging field, it is still unable to accurately measure gait speed and detailed spatiotemporal parameters in non-laboratory settings—that is, in real-world environments. It also requires further clinical trials to validate its clinical application. Future work should focus on validating spatio-temporal gait metrics and optimizing sensor configuration for reliable use in prevention and therapy.

3.1.1.3. Surface peripheral vascular monitoring

Local vascular impairment can lead to ulcers, necrosis, and eventual amputation. Many patients remain asymptomatic in early stages, delaying treatment until ulcers become severe. Wearable devices with integrated temperature and pressure sensors now allow non-invasive, real-time monitoring of superficial vascular status, supporting early detection and timely intervention to improve clinical prognosis.

Siren’s Neurofabric™ smart socks incorporate temperature sensors to enable real-time foot monitoring and direct data transmission to clinicians,^29,30 offering a convenient and cost-effective solution for remote diabetic care. Early detection of temperature variations can help identify ulcer risk before complications arise. Studies, such as one by Alexander M. Reyzelman et al., have shown significant temperature differences at multiple foot sites in diabetic patients with ulcers. They used this wearable device to predict the onset and progression of foot injuries. ³⁰ However, further research is needed to determine the specific threshold for predicting ulcers. While a 4°F differential has been proposed for prediction—achieving 97% sensitivity—it yields a high false positive rate of 57%.^31,32 Further research is needed to establish reliable thresholds through larger samples and longer monitoring periods. Additionally, the limited durability and single-use nature of the fabric present practical challenges for long-term use.

Smart insoles complement smart socks in predicting diabetic foot ulcers by integrating pressure and temperature sensors. These devices typically include a multiplexer, microcontroller, power supply, and communication module to collect and transmit plantar data for remote monitoring. ³³ Multiple studies have validated the feasibility of using smart foot pads to predict the early onset of ulcers.^34–36 While temperature differentials remain a primary predictive indicator, walking-induced variations have led some designs to incorporate pressure sensors for improved accuracy. Certain advanced systems also integrate real-time pressure feedback to actively correct gait and loading patterns. ³³ Although current devices rely heavily on temperature thresholds, future directions point to multi-parameter prediction combined with intervention capabilities. The integration of artificial intelligence is expected to enhance predictive accuracy and enable more personalized ulcer prevention strategies.

It is worth noting that most predictions are based on the temperature difference between the two legs. Some patients with only one leg may have difficulty using wearable devices. Lawrence A. Lavery et al. conducted an experiment showing that continuous daily monitoring of one leg can improve outcomes to a certain extent. ³⁷

3.1.1.4. Sophisticated wearable system(SWS)

The advancement of wearable technology has enabled the development of SWS, which integrate advanced materials and multi-sensor systems to continuously monitor a range of physiological indicators. These systems allow systematic data collection, transmission, and analysis, making home-based healthcare more portable, sustainable, and effective for disease prevention, progression monitoring, and intervention. Commercial examples such as smartwatches and fitness trackers exemplify this growing trend.

Some researchers have developed SWS targeting common lower-limb vascular diseases. ³⁸ By integrating sensor data with AI-driven storage and analytics, these systems offer visualized health feedback for both physicians and patients, significantly advancing home-based monitoring and early intervention. However, several challenges remain: no clear evidence to suggest a correlation with improved clinical outcomes; merging multiple technologies and harmonizing hardware with software present design difficulties; data security and patient privacy require robust safeguards; material selection and multi-system interference must be carefully addressed; and AI models demand large, clinically-representative datasets for reliable training and validation. ³⁹

3.1.2. Disease treatment intervention

3.1.2.1. Intelligent treatment

Wound management in venous diseases remains a key challenge in vascular surgery, often relying on subjective measures like visual inspection of size, exudate, and tissue quality—requiring frequent dressing removal. To enable objective and continuous monitoring, quasi-unipolar bioimpedance sensors have been developed. ⁴⁰ Bioimpedance measures the passive electrical properties of biological tissues. When skin is wounded—especially with damage to the stratum corneum and deeper layers—local impedance decreases due to fluid accumulation and structural disruption. As healing progresses, impedance gradually returns to baseline, providing a quantifiable indicator of tissue recovery. Therefore, the multi-electrode arrays detect reduced electrical impedance at wound sites, which normalizes during healing, allowing non-invasive assessment under dressings. Validation against acetate tracers shows strong correlation in healing progression. ⁴¹ Additionally, such electrodes can generate low-level H₂O₂ via direct current, providing antibacterial action and supporting healing. ⁴⁰ Moisture sensors have also demonstrated utility in managing exudate. ⁴² Combining these technologies could lead to integrated “smart dressings” that simultaneously monitor, treat, and simplify clinical care—though further trials are needed to confirm efficacy and support translation into practice.

In addition to smart dressings, so-called smart bandages have shown promise in mouse studies for treating chronic wounds. ⁸ These programmable devices use polymer microneedle arrays (MNAs) to deliver drugs or growth factors directly into deeper wound layers with minimal pain and inflammation. Drug release profiles can be tailored via a mobile application, supporting complete healing with reduced scarring. However, current systems face high costs and material biocompatibility challenges. Thus far, evidence is limited to animal models, and human trials are needed to evaluate safety, efficacy, and practical applicability.

3.1.2.2. Rehabilitation support

Virtual reality (VR) and augmented reality (AR) are emerging immersive technologies with growing medical applications. VR enables full engagement in simulated environments, while AR overlays digital elements onto the physical world. These tools can reduce perioperative anesthetic use by 25–75% through distraction, alleviate anxiety and pain in hospitalized patients, and support physical rehabilitation via gamified activities like virtual walking or table tennis.^43–46 However, clinical adoption remains limited by a lack of robust evidence and unresolved ethical concerns, including risks of addiction, reality substitution, and privacy infringement. Further validation is needed to establish efficacy and safety in medical contexts.

3.1.3. Doctor’s medical work support

Beyond patient-facing applications, wearable and digital technologies offer substantial advantages for medical professionals. AI supports clinical decision-making in diagnosis, treatment selection, and early intervention; robotics enhance surgical precision and reduce complications; and AR overlays anatomical guidance during procedures, improving spatial orientation and safety. ⁴⁷ Both AR and VR are also accelerating surgical training through immersive simulation.^43–45 As adoption increases, key challenges must be addressed: safeguarding patient data privacy, resolving technical interoperability, integrating innovations into existing clinical workflows, and evaluating the time and financial investments required for sustainable implementation.

There are already examples of commercial products being used in this area (such as Apple Glasses), but their application remains largely experimental, and there is a lack of definitive, large-scale, robust evidence demonstrating the extent to which they lead to quantitative improvements in clinical outcomes and training effectiveness.

4. Progress and prospects of wearable devices

4.1. Data algorithms

4.1.1. Artificial intelligence

AI has become a central focus in wearable technology, with advances in deep learning and convolutional neural networks enabling more flexible and powerful analysis of complex physiological data. These models excel in detecting cardiovascular conditions—particularly atrial fibrillation (AF)—where deep learning approaches have been shown to surpass traditional machine learning in accuracy. ⁷ Integrated into wearables, AI can subtype AF, assess individual risk, predict procedural outcomes, and guide treatment decisions in real time, contributing to more personalized and effective patient management. ⁴⁸

Nevertheless, AI models trained on data from clinical-grade hospital equipment still outperform those using wearable-derived data. ⁷ This gap stems from the predominant use of controlled, in-hospital datasets—rather than real-world wearable data—in developing and validating most algorithms. To enhance practical accuracy, future models must incorporate more diverse and large-scale data from wearable devices. Progress in predicting other conditions, such as sleep apnea or heart failure, has been hampered by limited public datasets and reliance on proprietary sources, leading to inconsistent algorithm performance and generalizability.

Current research often relies on single-signal analysis, but multimodal models integrating diverse biological signals, imaging data, and clinical information show superior potential. For instance, combining ECG with video-based facial blood flow analysis has improved atrial fibrillation detection accuracy. Multimodal integration represents a critical direction for enhancing the clinical utility of wearable devices. ⁴⁹

Moreover, practical tools for assessing real-world utility and bias risks in AI-assisted medicine remain underdeveloped. The “black box” nature of many deep learning models also limits clinical trust, as healthcare providers require interpretable reasoning behind AI-driven decisions. Enhancing model transparency—for instance, by incorporating explainable features such as heart rate variability between inputs and outcomes—could improve both acceptance and actionable insight.

4.1.2. PPG’s cutting-edge research

PPG technology holds significant potential for wearable applications in vascular surgery. Current research focuses on advanced waveform analysis and multimodal integration to enhance monitoring accuracy and clinical utility.

Mohamed Elgendi et al. identified 29 time-dependent PPG waveform features relevant to blood pressure estimation, ⁵⁰ highlighting the second-derivative ratios b/a and d/a as the most consistent predictors of systolic pressure. Combining these features with machine learning significantly improves prediction accuracy. A key limitation remains the dependency on calibration against standard devices like sphygmomanometers.

Henrik Hellqvist et al. demonstrated that finger PPG combined with machine learning can estimate aortic stiffness, identifying a novel early-waveform amplitude ratio as the strongest predictor, highly correlated with carotid-femoral pulse wave velocity (cfPWV) and aortic pulse wave velocity (aoPWV) (r = -0.81 and -0.75, p < 0.001). ⁵¹ However, the small sample size (n=33) and PPG’s inherent sensitivity to skin tone and obesity limit generalizability. Another study proposed the Katz fractal dimension (KFD) as a robust, noise-resistant arterial compliance (AC) metric derived from PPG, integrating compliance dynamics and vascular complexity. ⁵² While KFD correlated well with AC in virtual simulations (r=0.75), real-world validation using gold-standard measures like aoPWV or invasive tests remains essential. Future efforts should focus on larger clinical cohorts, standardized AC assessment, and integrating temporal patterns—such as the overnight stiffness index (OSI) ⁵³ —or AI-enhanced modeling to improve robustness and clinical applicability.

4.1.3. Others

Some studies integrate arterial viscoelasticity into one-dimensional elasticity models to analyze transient wave behavior. ⁵⁴ This approach shows potential for precisely localizing and tracking aneurysms and stenoses, enabling early detection of asymptomatic vascular conditions. However, linear models are limited to approximating routine cases and cannot accurately capture complex hemodynamic phenomena such as vortices or recirculation zones. Nonlinear models are essential for analyzing vascular abnormalities within intricate flow structures and their interactions with the arterial wall.

Building on existing dynamic blood pressure monitors using peripheral arterial tonometry, recent studies indicate these devices can also track parameters related to arterial stiffness—a clinically valuable but traditionally complex metric requiring specialized assessment. ⁵⁵ This capability positions wearable technology as a promising tool for continuous stiffness monitoring. Nevertheless, broader validation through large-scale studies and direct comparisons between normotensive and hypertensive populations remain essential to establish reliability and clinical relevance.

4.2. Technology development

4.2.1. Ultrasonic technology

Ultrasound offers notable advantages for vascular monitoring, including accuracy, portability, and deep tissue penetration. Wearable ultrasound devices show potential for non-invasive, continuous blood pressure estimation, potentially overcoming limitations of cuff-based methods. Integration with flexible patches could further enhance practicality, though current systems still face challenges in power consumption, circuit complexity, cost, and miniaturization. Three main ultrasound transducer technologies—piezoelectric(PZT), piezoelectric micro-electromechanical ultrasound transducers(PMUT), and capacitive micro-electromechanical ultrasound transducers(CMUT)—each present distinct trade-offs in performance and integrability. ⁵⁶ While dual-modal systems combining ultrasound and PPG have been explored,^57,58 broader multimodal integration remains limited. Key barriers include sensor stability during movement, data privacy, transmission reliability, and a lack of large-scale standardized clinical datasets needed to validate and standardize ultrasound-derived blood pressure measurements.

Cyprien Blanquart et al. developed an ultrasound-based method using a CMUT array for 3D microvascular imaging. ⁵⁹ The sensor successfully captured volumetric blood flow in volunteers’ fingers, wrists, and toes, maintaining consistent performance over months. It detected hemodynamic responses to temperature changes, systemic vasodilation, and physiological variations, enabling assessment of cardiovascular metrics such as heart rate, PWV, and pulsations. The system also supported software-assisted repositioning over six months, ensuring data consistency in long-term monitoring. Integration into wearables could enhance early detection and understanding of microvascular diseases. However, challenges such as motion artifacts, limited depth sensitivity, and the impracticality of traditional coupling agents remain. Flexible patches may offer a feasible alternative for wearable integration.

4.2.2. Flexible electronic materials

Flexible electronic materials are essential for wearable devices to conform to dynamic, non-rigid body surfaces. These materials leverage mechanical properties such as the piezoelectric effect, ⁶⁰ enabling applications like skin-attached or implantable pressure sensors for continuous arterial monitoring. ⁶¹ Multi-sensor arrays can be deployed across vascular sites to capture bidirectional pulse waves, offering a holistic view of vascular health unmatched by single-point sensing. ⁶² However, key challenges include ensuring biological safety—especially for implants—where risks of contamination or immune response must be addressed. Long-term accuracy, signal drift, and stability in real-world conditions also require further development (Figure 2).

Figure 2.

Technological evolution and integration pathways of wearable devices in cardiovascular care.

5. Clinical trial evidence and validation studies

5.1. Atrial fibrillation detection: From screening to outcomes

The Apple Heart Study (n=419,297) demonstrated that irregular pulse notifications via photoplethysmography had a positive predictive value (PPV) of 84% for AF confirmed by ECG patch monitoring. ⁶³ However, only 0.52% of participants received notifications, suggesting limited yield in unselected populations. In addition, the Fitbit Heart Study (n=455,699) reported AF detection in approximately 1% of participants, with 98% sensitivity for concurrent ECG-confirmed AF. ⁶⁴

The mSToPS trial randomized 2,659 participants to immediate versus delayed monitoring with a patch-based ECG monitor (Zio patch). ⁶⁵ At 4 months, AF was diagnosed in 3.9% of the immediate monitoring group versus 0.9% in the delayed group (HR 4.0, 95% CI 2.5-6.4). However, the trial was not powered to assess clinical outcomes such as stroke reduction.

The HEARTLINE study (NCT04276441), a pragmatic RCT enrolling 150,000 Medicare beneficiaries, aims to determine whether Apple Watch detection of AF reduces thromboembolic events. ⁶⁶ This trial addresses the critical evidence gap between AF detection and hard clinical outcomes.

While wearables show promise for AF screening in high-risk asymptomatic populations (e.g., age ≥65 with additional stroke risk factors), positive findings necessitate confirmation via standard ECG. A negative result does not exclude paroxysmal AF. The effect on stroke prevention remains to be established by ongoing RCTs including HEARTLINE.

5.2. Blood pressure monitoring: Accuracy and utility

Multiple validation studies have assessed cuffless BP devices against ambulatory blood pressure monitoring (ABPM). The Omron HeartGuide, a smartwatch with oscillometric cuff, achieved mean differences of -0.5±8.1 mmHg for systolic and -0.4±6.2 mmHg for diastolic BP compared to ABPM, meeting ISO 81060-2 standards. ⁶⁷

Despite technological advances, several limitations persist: (1) Most validation studies are short-term (<24 hours) in controlled settings; (2) Individual calibration remains necessary for many devices; (3) Performance may degrade with movement, posture changes, or in hypertensive patients; (4) No cuffless device is currently recommended for hypertension diagnosis by major guidelines. ⁶⁸

5.3. Heart failure and chronic disease management

The LINK-HF study demonstrated that a wearable patch measuring ECG, accelerometry, and thoracic impedance could predict heart failure hospitalization with 76% sensitivity up to 10 days before admission. ⁶⁹ In a case-control study evaluating Siren Socks, Reyzelman et al. compared 5 patients who developed foot injuries with 26 matched controls. ³⁰ The injury group demonstrated a significantly higher mean temperature differential (3.6°F, SD 3.0) compared to controls (2.2°F, SD 2.5), with a difference of 1.4°F (P<.001). However, the study’s small sample size (only 5 injury cases) limits generalizability, and further research is needed to establish optimal temperature thresholds and reduce false-positive rates. The summary table is shown below (Table 3).

Table 3.

Key clinical studies of wearable devices in cardiovascular disease.

Study/Device	Design	Population (n)	Primary endpoint	Key findings	Evidence level*	Clinical implications
Apple Heart Study (Apple Watch)	Prospective obs.	419,297	Irregular pulse notification	0.52% notified; PPV 84% for AF	2b	Screening tool for AF; requires confirmatory testing
Fitbit Heart Study	Prospective cohort	455,699	AF detection	1% AF detected; sensitivity 98%	2b	Similar utility to Apple Watch for AF screening
mSToPS Trial (Zio patch)	Pragmatic RCT	2,659	AF diagnosis at 4 months	Active monitoring: 3.9% AF vs usual care: 0.9% (HR 4.0)	1b	Increases AF detection; unclear impact on outcomes
HEARTLINE (Apple Watch)	Ongoing RCT	150,000 (planned)	Time to AF diagnosis, stroke	Results expected 2025	N/A (ongoing)	Will provide critical outcome data
Omron HeartGuide validation	Diagnostic accuracy	310	BP accuracy vs ABPM	Mean diff: -0.5±8.1 mmHg (SBP), -0.4±6.2 mmHg (DBP)	2a	Meets ISO standards; requires validation in diverse populations
LINK-HF (multiparameter patch)	Observational	100	HF hospitalization prediction	76% sensitivity up to 10 days pre-admission	2b	Early warning system; needs prospective validation
Smart Sock Temperature Monitoring (Siren Socks)	Case-control	5 injury cases + 26 controls	Temperature differential >4°F	Injury group: mean 3.6°F (SD 3.0)	3b	Continuous temperature monitoring may predict DFU; needs larger validation
				Control group: mean 2.2°F (SD 2.5)
				Difference: 1.4°F (P<.001)

^*Evidence Level: 1b = Individual RCT, 2a = Systematic review of cohort studies, 2b = Individual cohort study, 3a = Systematic review of case-control studies, 3b = Individual case-control study, 4 = Case-series.

5.4. Evidence gaps and future research priorities

A major gap exists between proven diagnostic accuracy and demonstrated improvements in patient-centered outcomes. There is insufficient evidence from large-scale RCTs showing that wearable-based AF screening reduces stroke incidence or that HF early-warning systems lower hospitalization rates. Future research must prioritize outcome-driven trials to establish whether detection translates to meaningful clinical benefit.

Secondly, study populations are predominantly young, healthy, and socioeconomically advantaged. Data are especially lacking for older adults with comorbidities, racial/ethnic minorities (where skin tone may affect optical sensors), and low-resource populations. Equity-focused research is needed to ensure broad applicability and prevent widening health disparities.

Thirdly, evidence on sustained engagement beyond 6–12 months is scarce. Factors such as comfort, charging burden, privacy concerns, and “wearable fatigue” limit real-world utility. Research should investigate strategies for long-term adherence and evaluate effectiveness in routine, uncontrolled settings.

Finally, Key barriers to implementation remain unstudied: lack of actionable data interpretation, undefined clinical pathways for responding to alerts, and unclear reimbursement models. Studies on workflow integration and health economics are urgently needed. The field lacks comparative effectiveness studies between devices, standardized validation methods, and unified data formats, creating interoperability barriers. Developing consensus standards and conducting head-to-head trials are essential for informed adoption (Table 4).

Table 4.

Research priority matrix.

Priority	Research domain	Key questions	Recommended study designs
High	Clinical Outcomes	Does screening/monitoring improve hard endpoints (e.g., stroke, HF hospitalization)?	Large-scale pragmatic RCTs
High	Health Equity	How do devices perform in diverse populations (e.g., elderly, ethnic minorities)?	Prospective diverse cohort studies
Medium	Health Economics	Is wearable monitoring cost-effective compared to usual care?	Cost-effectiveness analysis, budget impact modeling
Medium	Clinical Integration	How to optimize alerts, integrate into workflows, and design reimbursement models?	Hybrid implementation science studies
Low	Technological Refinement	How to improve sensor accuracy, comfort, and battery life?	Proof-of-concept, engineering optimization studies

6. Discussion

Over the past five years, wearable devices have been applied in cardiovascular surgery across two main domains: patient-facing and clinician-facing applications. The patient-focused segment is the most prominent, encompassing disease prediction and diagnosis (e.g., AF detection ⁷⁰ ), therapeutic intervention (e.g., smart wound dressings^71–73), and long-term physiological tracking (e.g., heart rate, gait analysis ⁷⁴ ). For healthcare professionals, these technologies aid in technical training and preoperative planning. Most devices leverage PPG technology, augmented with multi-sensor integration and AI, to enhance functionality and acziocuracy. The adoption of flexible electronics further supports device miniaturization and wearability.

Interdisciplinary integration has become a major catalyst in the development of wearable technology. AI algorithms enhance disease monitoring and early warning capabilities by addressing challenges such as PPG signal noise and improving predictive accuracy—especially when combined with large-scale models, as seen in atrial fibrillation detection.^75–77 Advances in materials science are equally critical, enabling miniaturized, flexible, and biologically safe sensors that support non-invasive, long-term wearability. ⁷⁸ Furthermore, growing clinical demand is shaping innovation: patients seek convenient home-based monitoring to replace certain hospital visits, while the need for surgical training and risk reduction is accelerating the adoption of VR simulations and AR navigation in healthcare. ⁷⁹

Wearable devices still face significant limitations and challenges. PPG-based measurements of heart rate, blood pressure, and vascular stiffness remain susceptible to motion artifacts and skin tone variations—issues that may be mitigated through multimodal integration, such as with AI or interface sensor.^80–83 Smart therapeutic devices, like localized drug-delivery bandages, exhibit low clinical translation rates and require validation in large-scale trials. Many advanced studies also suffer from insufficiently diverse clinical samples. High costs, particularly for AR/VR systems, limit accessibility in smaller healthcare settings. Data security and interoperability pose additional hurdles: continuous collection of sensitive health information raises privacy concerns, while the lack of unified industry standards leads to fragmented data formats and limits cross-device compatibility. Furthermore, AI models trained on in-hospital data often underperform in real-world wearable contexts, hindering both development and practical adoption (Figure 3).

Figure 3.

Key challenges and corresponding mitigation strategies for wearable device implementation.

Cutting-edge advancements are shaping the next generation of wearable technology, with PPG evolution at the forefront. Efforts are directed toward refining waveform analysis and identifying clinically relevant features through advanced optical sensors—such as multi-wavelength PPG ⁸³ —to enhance signal quality and feature richness. Coupling these with real-time AI analytics can significantly boost device efficiency and diagnostic value. Ultrasound wearables also hold major potential, though miniaturization challenges like power consumption, probe size, and coupling methods must be overcome to enable precise cardiac and vascular monitoring. Meanwhile, improving AI interpretability is essential to overcome “black box” limitations and build clinical trust. Continued algorithm development remains a key driver in enhancing the accuracy and utility of wearable data capture and analysis.

Although multiple large-scale studies have demonstrated the feasibility of PPG-based AF detection, notable variability exists in reported performance metrics. The Apple Heart Study reported a PPV of 84% for irregular pulse notifications confirmed by subsequent ECG patch monitoring, whereas the Fitbit Heart Study reported 98% sensitivity for concurrent ECG-confirmed AF.^63,64 The Huawei Heart Study, conducted in a Chinese population, reported a PPV of 91.6%. ⁸⁴ These differences likely reflect variations in study populations, reference standards, follow-up duration, and device-specific algorithms. Importantly, all three studies demonstrated that while PPG-based screening can identify undiagnosed AF, false positives remain a problem, and none has yet proven that such screening reduces stroke incidence—a critical evidence gap that ongoing trials such as HEARTLINE (NCT04276441) aim to address.

Regarding cuffless blood pressure monitoring, the evidence is similarly conflicting. Several validation studies have shown that devices based on PPG or PAT comply with the ISO 81060-2 standard under controlled laboratory conditions. ⁶⁷ However, as demonstrated by the study by Zheng et al., their accuracy declines significantly in real-world settings due to factors such as physical activity, changes in body position, and food intake. They found that daytime monitoring is less reliable than nighttime measurements due to fluctuations in vascular tone and pre-ejection phase. ²² Furthermore, an opinion piece by Avolio et al. emphasized that current major hypertension guidelines do not recommend any cuffless devices for diagnostic or therapeutic decision-making, citing issues such as calibration requirements, interindividual variability, and a lack of outcome data. ²¹

A key contribution of this review is the clear distinction between technologies that are ready for clinical application and those still in the research phase. Technologies that have been clinically validated, received regulatory approval (FDA or CE certification), and are supported by evidence include: (1) atrial fibrillation detection using PPG-based consumer smartwatches (Apple Watch, Fitbit, Huawei Watch), which can serve as a screening tool but require confirmation via ECG; (2) single-lead ECG patches for continuous heart rhythm monitoring (Zio patch); (3) activity trackers to promote physical activity during cardiac rehabilitation; and (4) smart socks for remote monitoring of diabetic foot (Siren Socks).

In contrast, several technologies remain in the research phase and have not yet been incorporated into routine clinical care. These technologies include: cuffless blood pressure monitors for diagnostic purposes (pending results from clinical trials and standardized validation protocols); smart dressings and microneedle array dressings for wound healing (currently limited to animal studies or small-scale human proof-of-concept trials); and AR/VR-based surgical navigation systems (lacking large-scale validation and cost-effectiveness data). We summarize this distinction in Table 2 to provide clinicians with a practical reference for evidence-based decision-making.

Even for validated technologies, successfully integrating them into clinical workflows remains challenging. First, alert fatigue poses a significant risk. If wearable devices generate frequent false-alarm notifications—such as the 57% false-alarm rate reported when using a 4°F threshold for temperature-based prediction of diabetic foot ulcers—clinicians may become desensitized, potentially leading them to miss true positive events. ³² Establishing optimal, context-specific alert thresholds and accounting for individual patient baseline variability are priority areas for research.

Second, data overload could overwhelm healthcare providers. A single patient wearing multiple sensor patches can generate thousands of data points every day. Without intelligent data filtering and prioritization, clinicians would face the nearly impossible task of reviewing all raw data. Future systems must incorporate artificial intelligence to provide concise, actionable summaries—for example, highlighting only persistent anomalies or significant trends, rather than every single data point.

Third, the reimbursement model for remote monitoring services remains underdeveloped. Although the Centers for Medicare & Medicaid Services has introduced Current Procedural Terminology codes for remote physiological monitoring, coverage varies across different payers and indications. Clinicians and healthcare systems lack clear financial incentives to invest in the infrastructure, training, and personnel required for wearable-based care, resulting in slow adoption despite the clinical promise of these technologies.

From a regulatory perspective, the regulatory landscape is evolving but remains fragmented. The FDA Center for Digital Health Excellence has pioneered groundbreaking device certification and pre-certification programs for software-based medical devices. However, international standardization lags behind; devices approved by the FDA may face additional requirements when seeking CE marking in Europe or NMPA approval in China. Furthermore, post-market regulation of wearable devices remains limited. Unlike pharmaceuticals, where adverse drug reactions are systematically reported, device-related issues—such as skin irritation caused by electrodes or inaccurate readings in specific populations—are often underreported.

Patient acceptance is another key factor influencing actual outcomes. Factors affecting adoption include: the comfort and appearance of the devices (smart socks and chest patches must be suitable for daily wear); ease of use (older adults may encounter difficulties with smartphone pairing and data transmission); privacy concerns (continuous health data collection raises questions about data ownership and security); and perceived value (if patients do not perceive tangible health benefits, they will stop using the devices). A human-centered design approach, which involves patient participation in device development, is crucial for addressing these barriers.

Several limitations should be acknowledged. First, this is a narrative rather than a systematic review; although we followed a structured search strategy, we did not conduct a formal quality assessment or meta-analysis. Second, the search timeframe may have excluded important earlier studies, although we cited key historical works where relevant. Third, the rapid pace of technological innovation means that some of the devices discussed may have been improved or received new regulatory approvals since our literature search. Finally, publication bias may exist, as positive results are more likely to be published than negative or inconclusive results.

As an “auxiliary tool”, wearable devices have met clinical standards in areas such as heart rate monitoring, blood pressure estimation, and atrial fibrillation prediction, already replacing certain traditional clinical tasks.^85,86 Looking ahead, wearable technology in cardiovascular medicine is likely to evolve in three directions. First, outcome-based randomized controlled trials will determine whether screening and monitoring can translate into significant reductions in stroke, hospitalization rates, and mortality—the ultimate test of clinical value. Second, the multimodal integration of PPG, ECG, ultrasound, and clinical data, combined with artificial intelligence, may enable more accurate and personalized predictions than single-signal approaches. Third, closed-loop therapeutic systems—which not only detect abnormalities but also provide automated interventions—represent the cutting edge of wearable technology. For patients, wearables can enable personalized preventive care, enhance quality of life, and improve health self-management. For clinicians, they support optimized resource allocation, precision surgery, and improved outcomes, while also contributing to continuous professional development. Realizing this potential will require deeper collaboration among clinicians, engineers, and policymakers to drive technological refinement, validation, and scalable integration into healthcare systems.

7. Conclusions

Wearable devices have already transformed certain aspects of cardiovascular care, particularly atrial fibrillation screening and physical activity monitoring. However, many promising technologies remain in the research phase, and successful clinical integration requires addressing challenges related to the quality of evidence, workflow design, regulatory clarity, reimbursement, and patient engagement. By systematically evaluating the existing evidence and distinguishing between technologies that are ready for application and those that remain experimental, this review aims to guide clinicians, researchers, and policymakers in decision-making within this rapidly evolving field.

Appendix.

List of abbreviations

CVD

Cardiovascular disease

AI

Artificial intelligence

DALYs

disability-adjusted life-years

CHD

Coronary Heart Disease

HF

Heart failure

HR

Heart rate

ECG

Electrocardiogram

PPG

Photoplethysmography

HRV

Heart rate variability

BP

Blood pressure

PWV

Pulse wave velocity

ML

Machine learning

PAT

Pulse arrival time

PPT

Pulse transit time

SBP

Systolic blood pressure

DBP

Diastolic blood pressure

PEP

Pre-ejection period

PF

Pattern fusion

IMU

Inertial measurement unit

SWS

Smart Wearable System

MNAs

Micro-needle arrays

VR

Virtual reality

AR

Augmented reality

AF

Atrial fibrillation

cfPWV

Carotid-femoral pulse wave velocity

aoPWV

Aortic pulse wave velocity

KFD

Katz fractal dimension

AC

Arterial compliance

OSI

Overnight stiffness index

PZT

Piezoelectric

PMUT

Piezoelectric micro-electromechanical ultrasound transducers

CMUT

Capacitive micro-electromechanical ultrasound transducers

RCTs

Randomized controlled trials.

ORCID iD

Qian Yang https://orcid.org/0009-0009-2444-3600