Artificial intelligence-based neonatal heart rate monitoring technologies: Systematic review

Abstract

BACKGROUND

Neonatal heart rate (HR) is an important parameter in the evaluation of newborn health and viability in the immediate postnatal period.

AIM

To evaluate the accuracy, reliability, and clinical applicability of emerging non-contact and artificial intelligence (AI)-assisted HR monitoring technologies in neonates compared to conventional electrocardiography (ECG)-based systems.

METHODS

A comprehensive literature search was conducted across PubMed, EMBASE, Google Scholar, and Cochrane databases from January 2013 through June 2025 following PRISMA guidelines.

RESULTS

The analysis revealed a progressive shift from contact-based ECG and pulse oximetry to camera-based photoplethysmography, thermal imaging, and AI-enhanced multimodal systems. These newer methods demonstrated a strong correlation with ECG readings, rapid signal acquisition, and improved robustness against motion and lighting variability.

CONCLUSION

Emerging non-contact, AI-assisted HR monitoring technologies offer accurate, safe, and efficient alternatives for neonatal care, supporting faster clinical decisions and improved outcomes. Future multicenter studies are required to validate accuracy and confirm clinical utility before routine clinical implementation.

Key Words: Artificial intelligence; Electrocardiography; Neonatal heart rate; Non-contact monitoring; Photoplethysmography

Core Tip: Electrocardiography (ECG) remains the reference standard for neonatal heart rate (HR) measurement, providing the fastest and most accurate detection during resuscitation. Limitations include electrode placement delays, skin fragility, and motion artifacts. Existing contact-based methods such as auscultation, palpation, and pulse oximetry are widely used but have limitations such as reduced reliability and dependency on perfusion, especially in preterm or unstable neonates. Emerging non-contact and artificial intelligence (AI)-enhanced technologies, including camera-based photoplethysmography, thermal imaging, Doppler, and multimodal AI systems, demonstrate promising accuracy compared with ECG,with faster signal acquisition and improved robustness against motion and lighting variability. This systematic review (2013 –2025) highlights the shift toward safe, contact-free, real-time HR monitoring while noting limited evidence during critical neonatal resuscitation scenarios.

INTRODUCTION

Heart rate (HR) has been identified as the optimal clinical index used to quantify intrauterine-to-extrauterine life transition efficacy and is also a significant parameter in the determination of neonatal resuscitation and stabilization interventions[1-5]. According to international standards, a heart rate below 100 beats/min indicates the need to assess ventilation and provide appropriate resuscitation. The timing and accuracy of neonatal HR measurement are critical, as they guide the assessment of the neonate’s need for resuscitation, the effectiveness of interventions, and overall clinical status. Currently, HR assessment in the delivery room is commonly performed using methods such as cardiac auscultation, palpation, pulse oximetry, and electrocardiography (ECG)[6-10]. Among these, ECG is considered the most accurate and rapid, while auscultation and palpation, although still in routine use, have been shown to be less reliable due to subjectivity and interobserver variability, particularly during emergencies[11].

Accurate HR assessment is crucial, as overestimation may delay necessary interventions, while underestimation can lead to unnecessary treatments, increasing the risk of future complications[12-14]. This highlights the need for more precise and standardized methods for HR measurement to improve patient outcomes. The selection of the HR measurement method is important because traditional methods such as auscultation and palpation are susceptible to subjective error and interobserver variation.

Although pulse oximetry is popular in clinical use, it is not without its limitations, such as signal delay and peripheral perfusion dependency, which can be compromised in ill or preterm neonates. By contrast, ECG offers a more immediate and accurate HR assessment and is an increasingly popular modality for monitoring in the delivery room[15]. In addition, recent research suggests that the threshold for HR intervention may vary across neonatal subgroups depending on cord management technique, mode of delivery, and gestational age, thereby challenging the traditional cutoff of 100 bpm. For example, physiologically based cord clamping may influence early HR recordings by sustaining placental transfusion and promoting cardiovascular stability during the transition to extrauterine life[16-20].

Since HR is closely associated with systemic oxygenation, changes in the use of supplemental oxygen in preterm and term neonates may influence the expected pattern and prognostic significance of HR during the first minutes after birth[21]. A pulse oximeter measures oxygen saturation and HR but with a delay compared with dry-electrode or newer-generation ECG systems, which provide real-time data. As precision medicine continues to take center stage in neonatology, additional studies will be necessary to improve HR evaluation methods and set the best intervention thresholds for defined neonatal subgroups[22]. The objective of this review was to summarize recent evidence regarding factors affecting HR, as well as measurement techniques, in newborns. The review also provides an overview of recent research on HR as a predictor of outcome in various newborn subgroups, highlighting the significance of precise HR determination in the management of neonatal resuscitation and better clinical outcomes. A pulse oximeter uses light absorption at two different wavelengths to detect peripheral pulse wave and estimate HR. Its accuracy depends on adequate peripheral perfusion and can be affected by motion.

MATERIALS AND METHODS

This systematic review protocol was developed in accordance with PRISMA guidelines. However, it was not registered in a public registry such as PROSPERO.

The review process followed the PRISMA guidelines, which provide a structured and comprehensive framework for the identification, selection, evaluation, and synthesis of relevant studies. This methodological approach not only enhances reproducibility and methodological rigor but also strengthens the credibility of the conclusions drawn from the review.

Population, intervention, comparator, and outcome framework: Inclusion and exclusion criteria

All published studies evaluating HR assessment technologies in neonates (0-28 days after birth), including both term and preterm neonates, were included. Eligible studies encompassed investigations conducted in delivery rooms, neonatal intensive care units (NICUs), and during neonatal resuscitation, addressing both term and preterm neonates. The technologies considered included digital stethoscopes, smartphone tap-based applications, pulse oximetry, ECG, Doppler ultrasound, photoplethysmography (PPG), camera-based PPG, and sensor-based techniques. Both in vivo studies and simulation-based evaluations were included, provided that they directly assessed HR measurement accuracy, latency, or efficacy. No restrictions were imposed concerning language, publication date, or study setting. However, studies conducted exclusively on pediatric or adult populations were excluded. Similarly, studies that used these technologies for parameters other than HR monitoring (such as oxygen saturation measurement using pulse oximetry or respiratory rate monitoring via PPG) were excluded from this review.

Search strategy and study selection

The search primarily focused on English-language publications; however, non-English studies were considered when sufficient methodological information was available. A comprehensive literature search was conducted across multiple electronic databases, including PubMed, Google Scholar, EMBASE, and the Cochrane Central Register of Controlled Trials, covering the period from January 2013 through June 2025. A combination of controlled vocabulary terms and free-text keywords related to “neonates”, “heart rate monitoring”, “digital stethoscope”, “pulse oximetry”, “ECG”, “photoplethysmography”, “smartphone-based HR monitoring”, and “Doppler ultrasound” was used. Boolean operators, truncations, and wildcards were applied to enhance search sensitivity. Additionally, manual screening of reference lists from relevant articles was performed to capture additional studies not indexed in the primary databases. The study selection was carried out independently by two reviewers, and any discrepancies in inclusion decisions were resolved through discussion with a third reviewer.

Data extraction

Data extraction was performed independently by two investigators using a standardized Microsoft Excel data collection form. Extracted data included study design, sample size, study population (term and preterm neonates), technology type, setting (delivery room or NICU), outcome measures (HR accuracy, time-to-detection, reliability), and comparator methods (commonly ECG). When necessary, corresponding authors were contacted for clarification or to obtain missing information. The extracted parameters also encompassed performance indicators such as signal acquisition time, mean absolute error (MAE), latency, rate of signal dropout, and correlation coefficients relative to ECG or pulse oximetry. Disagreements during data extraction were resolved by consensus or consultation with a third investigator.

Risk of bias assessment

Two independent reviewers assessed the methodological quality and risk of bias for each included study using standardized evaluation frameworks. For randomized controlled trials, the Cochrane Collaboration’s Risk of Bias Tool was applied. For diagnostic accuracy studies, the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool was used. Artificial intelligence (AI)-based prediction model studies were evaluated using the PROBAST tool. Any disagreements between reviewers were resolved through consensus. The risk of bias findings were summarized in a structured format, categorizing each study as low, moderate, or high risk to facilitate comparative interpretation across the different technologies.

Data analysis and synthesis

A narrative synthesis approach was adopted to summarize the performance of different HR assessment technologies used in neonates. No quantitative meta-analysis was performed because of clinical and methodological heterogeneity among studies. Quantitative data were tabulated to describe accuracy, response time, and reliability compared with the gold standard ECG. Descriptive statistics, including measures of central tendency (mean or median) and dispersion (standard deviation or interquartile range), were extracted from individual studies and narratively synthesized.

Technologies such as digital stethoscopes and smartphone-based applications were analyzed for feasibility and latency; pulse oximetry and PPG-based methods were evaluated for HR precision and signal stability; Doppler ultrasound and sensor-based systems were assessed for real-time performance and applicability in low-resource settings. The synthesis also included comparative evaluation of the time to reliable HR acquisition after birth, which is particularly critical during neonatal resuscitation. Study-level results were summarized narratively and graphical representations were used where adequate quantitative data were available. Collectively, this analytical approach provided a comprehensive overview of the reliability, accuracy, and clinical utility of emerging digital and AI-assisted HR assessment technologies in neonatal care.

RESULTS

Characteristics of included studies

Across the included studies, comparative performance was assessed using reported correlation coefficients relative to ECG, MAE, signal acquisition latency, and feasibility under clinical conditions, where available.

A total of 16 eligible studies published between January 2013 and June 2025 were included in the systematic review. These studies collectively evaluated a range of non-contact and non-invasive technologies for HR monitoring in neonates (Figure 1). The research focused on improving the accuracy, reliability, and clinical usability of these techniques as alternatives to conventional ECG monitoring, particularly in the NICU setting.

Figure 1 PRISMA flow diagram illustrating the study selection process for the systematic review.

Figure 1 PRISMA diagram of the steps in the record review and selection process. Excluded studies were due to non-neonatal populations, absence of HR outcome data, duplicate records, use of technologies not related to HR monitoring, and insufficient methodological reporting.

Study characteristics

The included studies exhibit a progressive evolution of neonatal HR monitoring technologies (Table 1). Early studies (2013-2017) primarily validated the feasibility and basic accuracy of optical and Doppler systems, whereas recent studies (January 2020 to June 2025) emphasized AI-based integration, 3D imaging, and multimodal fusion for enhanced precision under clinical conditions. Sample sizes remain modest, reflecting the sensitive nature of neonatal trials, although methodological rigor varied across studies and publication years.

Table 1 Characteristics of the included studies.

No.	Ref.	Study design	Sample size	Technology used	Key findings
1	Mestha et al[16]	Observational feasibility study	Neonates in NICU	Webcam-based photoplethysmography	Demonstrated the feasibility of continuous pulse monitoring using a webcam in the NICU
2	Ruhrberg Estévez et al[30]	Prospective clinical study	18 neonates	RGB-D camera system	Enabled continuous non-contact monitoring of neonatal vital signs in intensive care settings
3	Chen et al[22]	Prospective experimental study	15 neonates	Motion-compensated camera-based heart rate detection	Improved the accuracy of heart rate measurement despite neonatal motion artifacts
4	Gangaram-Panday et al[19]	Observational validation study	25 neonates	Dynamic light scattering technology	Demonstrated a strong correlation with ECG-based heart rate monitoring
5	Peeples et al[25]	Comparative observational study	40 neonates	Handheld Doppler ultrasound	Provided accurate bedside heart rate assessment in preterm neonates
6	Nagy et al[20]	Algorithm development and validation study	30 premature neonates	Camera-based monitoring algorithm	Achieved reliable continuous heart rate tracking in NICU settings
7	Lake et al[21]	Retrospective analytical study	> 100 neonates	Heart rate variability bioinformatics	The heart rate characteristics index predicted neonatal sepsis risk
8	Abdou et al[23]	Proof-of-concept protocol study	Neonatal resuscitation model	Novel heart rate detector device	Proposed a device for early heart rate detection during neonatal resuscitation
9	Maurya et al[24]	Pilot observational study	Neonates	Thermal and visible imaging	Demonstrated the feasibility of non-contact monitoring using thermal and visible imaging
10	Ahmad Hatib et al[31]	Prospective clinical study	Pediatric patients including neonates	Remote photoplethysmography (rPPG)	Demonstrated the feasibility of camera-based pulse detection in pediatric care
11	Cobos-Torres et al[26]	Experimental feasibility study	Neonates	Photoplethysmography imaging	Demonstrated simple non-contact neonatal vital sign monitoring
12	van Gastel et al[32]	Experimental NICU feasibility study	Neonates	Non-contact optical pulse monitoring	Demonstrated heart rate monitoring under low-light NICU conditions
13	Huang et al[33]	Dataset development and AI modeling study	Neonatal dataset	Spatio-temporal neural networks	Developed benchmark dataset for non-contact neonatal heart rate monitoring
14	Rapczynski et al[6]	Experimental signal processing study	Laboratory validation	Camera-based heart rate signal processing	Demonstrated the influence of video encoding on heart rate estimation accuracy
15	Marchionni et al[15]	Experimental optical monitoring study	Preterm neonates	Optical respiratory and heart rate monitoring	Enabled simultaneous measurement of respiration and heart rate
16	Nukaya et al[17]	Engineering feasibility study	Neonates	Contact-free physiological monitoring system	Demonstrated monitoring of heart rate, respiration, and body movement using non-invasive sensors

HR: Heart rate; NICU: Neonatal intensive care unit; ECG: Electrocardiography; RR: Respiratory rate; RGB: Red, green, blue; rPPG: Remote photoplethysmography.

ECG

ECG served as the reference standard for HR measurement in nearly all studies. Lake et al[21] expanded the clinical utility of ECG through the HR characteristics (HRC) index, which helped identify early signs of neonatal sepsis by analyzing HR variability. Mestha et al[16] used ECG as a baseline comparator to validate camera-based photoplethysmography (PPG) in NICU neonates, demonstrating comparable HR measurement accuracy under steady illumination. These findings highlight ECG’s diagnostic and reference value in validating emerging HR monitoring technologies. However, the requirement for electrodes and continuous skin contact makes ECG less suitable for medically fragile preterm neonates.

Pulse oximetry

Pulse oximetry remains a routinely used, contact-based tool for HR measurement, primarily used as a reference or comparator in validation studies. Although it provides quick HR readings, several studies noted susceptibility to motion artifacts and reduced accuracy during low perfusion states. This has encouraged the development of camera-based, non-contact optical methods, which avoid direct skin-contact sensors. No standalone study among the included studies investigated pulse oximetry as a primary innovation, suggesting comparatively limited recent innovation.

Digital stethoscope and tap-based apps

Abdou et al[23] presented an innovative digital HR detector prototype for use during neonatal resuscitation, aiming to provide real-time HR monitoring and feedback without requiring conventional ECG electrodes. Such acoustic or touch-assisted monitoring technologies could improve delivery-room monitoring where rapid HR detection is critical. Although still under validation, these systems represent the next step toward portable and minimally contact-dependent neonatal monitoring technologies.

Thermal and dual-modality imaging techniques

Thermal imaging and dual-sensor modalities are emerging technologies in neonatal monitoring. Chen et al[22] evaluated RGB camera-based PPG for HR estimation in neonates. The algorithm showed high accuracy compared with ECG-derived HR values, even in the presence of body movements. Similarly, Maurya et al[24] used combined visible-spectrum and thermal imaging to simultaneously measure respiratory rate and HR, enhancing diagnostic integration. These techniques are beneficial in low-light NICU environments and reduce the risk of skin irritation associated with adhesive sensors, making them potentially suitable for preterm or sensitive neonates.

Doppler ultrasound

Peeples et al[25] demonstrated that handheld and rapid-response Doppler ultrasound devices offer real-time HR and hemodynamic assessment in neonates. Although slightly slower in response than ECG, Doppler systems are clinically useful in emergency and immediate postnatal care settings, particularly during resuscitation. Their key advantage lies in direct blood flow detection, providing physiologically relevant cardiovascular assessment; however, performance can be affected by probe positioning and operator experience.

Camera-based PPG and imaging systems

Camera-based HR estimation represents the most actively explored approach among the reviewed studies. Mestha et al[16] demonstrated the feasibility of webcam-based monitoring, while Nagy et al[20] further developed camera-based monitoring algorithms that enabled reliable HR detection in premature neonates. Chen et al[22] improved HR estimation accuracy through motion artifact reduction algorithms. Additional optical monitoring approaches have been investigated in neonatal settings[26-29]. Ruhrberg Estévez et al[30] validated 3D multimodal camera systems, showing stable readings even under variable lighting conditions. Further developments in imaging-based monitoring systems have also been reported[31]. Van Gastel et al[32] advanced low-light monitoring algorithms, supporting the development of continuous NICU-compatible monitoring systems. These findings collectively show that optical imaging is progressing from laboratory feasibility to potential clinical implementation, providing safe and effective alternatives to contact sensors.

PPG

PPG-based systems, including dynamic light scattering (DLS) -based methods and remote PPG (rPPG), demonstrated considerable potential as non-invasive HR measurement alternatives. Gangaram-Panday et al[19] reported strong agreement between DLS-derived HR and ECG readings, supporting its reliability. Recent advancements have expanded these technologies through a two-phase clinical trial, demonstrating rPPG’s high reliability and contactless potential for both pediatric and neonatal monitoring. The major benefit is elimination of direct skin contact, reducing infection risks and discomfort in neonates admitted to the NICU.

Sensor-based and AI-driven approaches

Advanced AI frameworks are being integrated with imaging systems to enhance performance under clinically variable conditions. Huang et al[33] developed a spatio-temporal neural networks using a large neonatal dataset, achieving superior HR prediction performance despite variations in motion and lighting conditions. Nagy et al[20] contributed similar algorithmic improvements for premature neonates. These findings indicate the growing role of machine learning in automating and refining neonatal monitoring systems for reliability and clinical scalability. Current AI models have not yet consistently achieved validation (± 2 bpm vs ECG), emphasizing the significance of large-scale, real-world testing before routine use. While early experimental datasets demonstrate promising correlation, large-scale clinical validation remains necessary before clinical equivalence can be established.

Risk of bias assessment

The methodological quality of the included studies was generally good, with most studies demonstrating a low risk of bias across key domains. As shown in Figure 2, selection bias, performance bias, and detection bias were low in over 85% of the studies due to appropriate randomization procedures, blinding, and standardized measurement tools. A small number of studies exhibited moderate attrition bias, primarily resulting from incomplete data or participant dropout during follow-up. Reporting bias appeared to be low, as nearly all studies reported outcomes consistent with their protocols. The overall risk-of-bias distribution across the included studies is shown in Figure 3.

Figure 2 Risk of bias summary. Domains evaluated included selection, performance, detection, attrition, and reporting bias. Orange, blue and green bars represent low, moderate, and high risk of bias, respectively.

Figure 3 Intended figure showing the classification of neonatal heart-rate monitoring technologies.

DISCUSSION

Role of AI in neonatal HR monitoring: AI enhances the accuracy of neonatal HR monitoring by correcting motion artifacts, normalizing illumination, and enabling continuous real-time analysis of imaging data. The results of this systematic review indicate a substantial technological evolution in neonatal HR monitoring, progressing from traditional contact-based modalities such as ECG and pulse oximetry toward non-contact, optical, and AI-driven systems. The comparative evaluation of these technologies revealed that accuracy, response time, and clinical applicability are the primary determinants for neonatal resuscitation and intensive care settings.

ECG continues to serve as the clinical reference standard for HR determination in neonates, as reaffirmed by multiple studies included in this review[16,22]. The 2015 International Consensus on Neonatal Resuscitation by Perlman et al[34] and Wyckoff et al[35] emphasized that ECG provides the fastest and most accurate HR readings, which are critical during the initial seconds of resuscitation. However, practical challenges such as delays in electrode placement, skin fragility, and motion artifacts limit its immediate utility in the delivery room as reported by Kamlin et al[36]; Luong et al[37]. Across the included studies, it was consistently observed that none of the emerging technologies surpassed or outperformed ECG in precision; however, several demonstrated, clinically acceptable accuracy under stable conditions along with faster signal acquisition.

In comparison, studies such as Mestha et al[16] and Nagy et al[20] demonstrated that camera-based PPG systems can achieve promising accuracy comparable to ECG under controlled conditions, with the advantage of non-contact operation. While ECG remains indispensable for the detection of pulseless electrical activity (PEA) and cardiac rhythm abnormalities, Myerburg et al[27] and Patel et al[28], non-contact systems may complement ECG by reducing intervention time in fragile preterm neonates. The Textbook of Neonatal Resuscitation[38] further supports the integration of ECG with emerging non-invasive modalities to balance diagnostic precision and patient comfort.

Pulse oximetry is widely used in neonatal units because of its simplicity and availability; however, studies in both the present review and prior clinical literature underline significant limitations. Van Vonderen et al[39] and O’Donnell et al[40] demonstrated that pulse oximetry tends to underestimate HR compared to ECG during the first minute after birth, and may require up to 30-45 seconds to establish a reliable reading.

The reviewed studies corroborate these findings, emphasizing that although pulse oximetry offers continuous monitoring, it performs poorly in low-perfusion states and during neonatal movement, conditions common during resuscitation. Swenson et al[41] recommended pulse oximetry as a routine adjunct but not as a primary HR assessment tool during resuscitation, because of delays in signal acquisition and dependence on peripheral perfusion. In contrast, PPG and camera-based imaging systems offer improved temporal resolution and greater tolerance to motion artifacts, as seen in studies by Gangaram-Panday et al[19]. These non-contact technologies may help overcome limitations associated with pulse oximetry, enabling faster and more stable HR readings in dynamic clinical environments.

Handheld and rapid Doppler ultrasound devices, such as those evaluated by Peeples et al[25], have been shown to be effective for real-time HR detection during neonatal care. Consistent with findings reported by Wyllie et al[42] and Perlman et al[43], Doppler methods are particularly beneficial in detecting HR when pulse or peripheral perfusion are difficult to assess clinically, such as in asphyxiated neonates. However, their use requires trained operators and may be limited by motion artifacts and probe alignment issues.

Traditional contact-based PPG has evolved significantly through dynamic light scattering (DLS) and remote photoplethysmography (rPPG) technologies. Studies by Gangaram-Panday et al[19] demonstrated that optical PPG systems exhibit strong correlation with ECG-derived HR values, supporting their potential use for continuous neonatal HR monitoring. Additionally, these optical systems offer motion-compensated and illumination-adjusted algorithms, reducing artifacts during patient movement, an issue that continues to challenge both ECG and pulse oximetry, as reported by Van Vonderen et al[39] and O’Donnell et al[40].

Among all reviewed modalities, camera-based PPG and imaging systems demonstrated the most significant clinical advancement. Studies by Mestha et al[16], Chen et al[22], and Ruhrberg Estévez et al[30] collectively validate the feasibility of accurate, contactless HR monitoring across diverse illumination and motion conditions. This progress directly addresses a critical limitation identified in neonatal resuscitation guidelines: The delay associated with, and invasiveness of, contact-based sensors, as highlighted by Perlman et al[34] and Wyllie et al[42].

Recent prototypes, such as those proposed by Abdou et al[23], explore digital stethoscopes and acoustic HR detection for use during neonatal resuscitation. In the context of PEA, accurate detection of mechanical cardiac activity remains essential, as emphasized by Mehta and Brady[44] and Sillers et al[29]. Digital stethoscope systems could complement ECG by differentiating electrical activity without mechanical contraction, thereby reducing the risk of misinterpretation during resuscitation, particularly in fragile or preterm neonates.

Clinical implications

The integration of non-contact and AI-enhanced HR monitoring systems aligns with the global shift in neonatal resuscitation toward minimizing invasive procedures and maximizing the accuracy of early intervention. The findings underscore that camera-based PPG and AI-assisted optical systems may complement, or in selected settings potentially reduce reliance on, contact-based sensors, especially in neonates requiring minimal-touch care. Future neonatal guidelines may incorporate hybrid monitoring frameworks, combining ECG for rhythm detection with optical or acoustic modalities for rapid HR confirmation, a model already envisioned in the 2015 and 2016 AHA/AAP updates.

Limitations and future prospects

As quantitative meta-analysis was not feasible because of heterogeneity and limited number of included studies, results should be interpreted with caution.

Although promising, the reviewed optical and AI-based systems remain limited by small sample sizes and predominantly controlled settings. Future studies should focus on large-scale validation, integration into resuscitation protocols, and real-time synchronization with ECG to ensure clinical reliability. The AI-driven fusion of camera, Doppler, and acoustic data may represent a potential future direction for automated neonatal resuscitation support systems, consistent with the direction outlined by the European Resuscitation Council and the AHA. This systematic review included only open-access literature, which represents a methodological limitation. This restriction may have reduced the overall comprehensiveness of the review.

CONCLUSION

This systematic review highlights a major transition in neonatal HR monitoring from traditional contact-based methods such as ECG and pulse oximetry toward advanced non-contact and AI-assisted systems. Emerging technologies, including camera-based PPG, thermal imaging, and Doppler ultrasound have demonstrated promising accuracy in early-phase studies along with rapid response times and improved comfort for fragile neonates. These innovations have the potential to reduce infection risk, enable continuous monitoring, and support earlier intervention during neonatal resuscitation. While ECG remains the gold standard, integrating optical and AI-driven systems may provide a complementary, safer, and more efficient monitoring approach. Future large-scale clinical validation studies are essential to standardize these technologies for routine neonatal intensive care and delivery room settings.