Introduction

When exposed to a cold environment, the sympathetic nervous system rapidly become excited, causing the human body to enter a state of stress. This results in a rapid heart rate, peripheral vascular constriction, and a significant increase in cardiac workload and oxygen consumption, which provokes an increased risk of arrhythmias, posing significant challenges for individuals. Moreover, prolonged cold exposure can induce alterations in cardiac structure, potentially causing myocardial hypertrophy and fibrosis. This, in turn, damages the heart’s conduction system and further impacts its electrical activity1,2,3. However, there is currently a scarcity of reliable indicators for screening related risks.

Ventricular repolarization (VR) heterogeneity within the myocardium is a complex electrical phenomenon and a crucial stage in cardiac electrical activity4. Perturbations of repolarization may favor the onset of lethal ventricular arrhythmias5,6,7. Thus, the assessment of VR has become an interesting and useful tool for the risk stratification of arrhythmia events8,9,10. Some of the most explored VR predictors in clinical practice are QT interval, the heart rate-corrected QT interval (QTc), and other recently published markers like Tpeak-Tend (TpTe), corrected TpTe interval (TpTec) by heart rate, and TpTe/QT ratio11,12. These markers have demonstrated a high level of usefulness in indicating patients at high risk for developing cardiac arrhythmias across various clinical conditions13,14,15.

The conventional 12-lead electrocardiogram (ECG) has long been a cornerstone in reflecting cardiac electrical activity. However, its complexity, poor portability, and the challenge of conducting it outside the hospital make it difficult to meet examination requirements in certain outdoor conditions16. With the widespread adoption of intelligent wearable medical devices, the development of mobile ECG signal acquisition devices has effectively mitigated these limitations are widely used in disease diagnosis17,18. However, there is a scarcity of reports on the use of wearable ECGs in extreme environments. Exposure to cold, particularly extreme cold, can cause muscle tension, leading to myoelectric interference. Additionally, cold stress results in the constriction of blood vessels on the body surface, slowing down blood circulation and affecting the conduction of the ECG signal, which in turn impacts the accuracy of the test results.

Therefore, we independently developed a 12-lead handheld wearable ECG. A multi-layered fabric electrode design, like a sandwich structure, was employed to obtain more stable and accurate ECG signals, to optimize applications in cold environments, and to achieve early identification of ECG abnormalities. Our retrospective study was conducted to compare the diagnostic accuracy of ECG obtained using the wearable device versus that from a conventional machine. Subsequently, we also investigated the changes in VR within the cohort after exposure to a extreme cold environment (-35 °C) and to identify the factors that could influence VR, with the hope of screening individuals at high risk of arrhythmia under this condition and reducing the occurrence of sudden cardiac death events.

Methods

Wearable ECG

A self-developed 12-lead handheld wearable ECG signal acquisition device was utilized (Registration Number: No. 20232070045; frequency response: 0.05–100 Hz, sampling frequency: 500 Hz, digital resolution: 12bits, input impedance ≥ 10 MΩ, standard sensitivity: 10 mm/mV, error ≤ ± 5%), which positioned the 10 electrodes required for a full-lead ECG, adhering to clinical measurement standards. All electrodes were integrated onto the palm side of the glove, as depicted in Fig. 1A. A specialized user interface facilitates real-time waveform visualization, and the captured signals can be wirelessly transmitted to an external display platform for immediate analysis and remote monitoring.

Fig. 1
figure 1

Schematics of wearable ECG electrode positions and sandwich-structure fabric electrode. Wearable ECG electrode placement schematic: V1: The electrode at the base of the thumb was positioned on the right margin of the sternum, between the 4th and 5th ribs; V2: The electrode, located equidistant between V1 and V3, was placed in the middle sternal line; V3: The electrode was situated at the intersection of the base of the index finger and the base of the middle finger; V4: The electrode was positioned at the first knuckle of the ring finger; V5: The electrode at the second knuckle of the ring finger was placed at the left 5th intercostal space; V6: The electrode at the third knuckle of the ring finger was also placed at the left 5th intercostal space; RA: The electrode at the tip of the thumb was positioned on the right chest wall, approximately at the lower margin of the second rib; LA: The electrode at the tip of the index finger was placed on the left chest wall, around the lower margin of the second rib; LL: The electrode at the tip of the little finger was positioned on the lower margin of the left chest wall; RL: The electrode was located on the lower right side of the hand, near the wrist (A). Multi-layer electrode configuration schematic (B). QT interval curve fitting plot (C). QT interval Bland-Altman plot (D). QT interval composite distribution (E). ECG waveform localization contrast (F). AI-based ECG waveform mapping performance (G).

Full size image

To optimize performance in cold environments, the electrode design incorporated fabric-based conductive materials combined with insulating layers. A novel sandwich-structured fabric electrode (Fig. 1B) has been employed, offering mechanical stability and consistent skin contact. The outermost electrode layer consisted of a flexible conductive material, which was connected to the underlying flexible printed circuit board (FPC) via ultra-fine wires. Additionally, a high-density foam pad was embedded beneath the FPC to improve contact compliance.

To mitigate temperature-induced performance drift, a real-time thermal compensation mechanism was integrated into the device’s system control algorithm. A negative temperature coefficient thermistor was employed to monitor ambient temperature, and its analog output was digitized using a high-precision analog-to-digital converter (AD7124). A combination of lookup tables and linear interpolation was used to compute real-time temperature values, which were then used to compensate for system-level parameters. Furthermore, a temperature-compensated crystal oscillator (TCXO; model X1G0054410203) was embedded to minimize frequency instability under variable environmental conditions. The outer casing adopted a snap-fit enclosure with three alignment points and two sliding locks, allowing for tool-free assembly and disassembly while maintaining a robust seal and ease of maintenance.

To mitigate discrepancies between wearable and standard ECG signals, a paired data acquisition protocol was established during device validation. Simultaneous recordings from both the wearable and standard ECG devices were collected and used to construct a synchronized training database. Leveraging this dataset, a deep regression network was developed, which integrates a stacked autoencoder and a bidirectional long short-term memory (Bi-LSTM) network, to achieve accurate signal transformation.

The model architecture comprised three modules: an encoder, a temporal sequence modeler, and a decoder. The encoder consisted of three fully connected layers with rectified linear unit activations, sequentially reducing the 500-point input heartbeat segment to a 128-dimensional latent representation. The Bi-LSTM module, with a hidden size of 128, was utilized to capture bidirectional temporal dependencies across sequential heartbeats. Finally, the decoder replicated the encoder’s structure to reconstruct the mapped waveform, restoring it to the original 500-point dimensionality and aligning it with the standard ECG waveform. Model training was conducted using the mean squared error loss function. The Adam optimizer was employed with an initial learning rate of 1e− 3. To prevent overfitting, a dropout layer (rate = 0.3) was inserted between the encoder and decoder. The dataset (n = 90780) was randomly split into training (80%) and validation (20%) sets. A 5-fold cross-validation strategy was implemented to evaluate model generalizability and the proposed architecture achieved a mean root mean square error of 0.09 mV relative to the reference ECG (Fig. 1C-G).

Additionally, the wearable ECG’s performance was tested in a simulated chamber with adjustable temperatures to evaluate its detectability. The results indicated that the wearable ECG retained functionality and accuracy even under cold environmental conditions (-40 °C).

Study populations and protocols

This study recruited 174 healthy young volunteers aged over 18 years. This study retrospectively compiled data from both before (room temperature @18 ± 2 °C) and after (temperature: -32~-36 °C, wind speed: 4.1 ~ 6.4 m/s, humidity: 65 ~ 82%, atmospheric pressure: 950 ~ 952 hPa) their exposure to an outdoor cold environment for 2 h in Jan 2024 in Heilongjiang province. The subjects uniformly donned winter attire for non-strenuous outdoor activities, and the main exposure to cold in the study was through the respiratory tract and the forehead skin. The exclusion criteria were as follows: (1) Cardiovascular diseases; (2) Skin diseases; (3) Mental illness; (4) Abnormal function of other organs; (5) Implanted pacemaker, cardioverter-defibrillator, or any electrical devices; (6) Recent medication use, whether terminated or not. Collected health information included demographics, physiological data, ECG, and echocardiography. The flow chart of the study is shown in Fig. 2.

Fig. 2
figure 2

Flow chart of the study.

Full size image

Demographics

Body mass index (BMI) was calculated by dividing body weight in kilograms by the square of the height in meters (kg/m2). Participants’ health information was collected including age, gender, smoking history, alcohol intake, race and living duration in cold zones. Body core temperature was measured by an ear thermometer (Omron TH839s) and percutaneous oxygen saturation (SpO2) measurements were done using a monitor (Omron HPO-201T) in all subjects. Blood pressure (BP) was measured by a sphygmomanometer (Redmond Spacelabs-90207) after at least 10 min of rest. Subjects remained still during the measurements and avoided unusual physical activities. The radial artery was cannulated for blood samples, and blood gas analyses were performed using an automatic blood gas analyzer (Abbott i-STAT 300-G).

Electrocardiographic examination

Within 5 min of the wearable ECG recording, a conventional 12-lead ECG was also conducted employing standard lead configurations at a speed of 25 mm/s (ECG-2250, Nihon Kohden, Tokyo, Japan; frequency response: 0.05–150 Hz, sampling frequency: 1000 Hz, digital resolution: 16bits, input impedance ≥ 50 MΩ, standard sensitivity: 10 mm/mV, error ≤ ± 5%). Electrocardiographic parameters were manually and blindly calculated using slide callipers and analyzed by two experienced cardiologists. The QRS duration, QT interval, and TpTe interval were measured from the most clearly defined wave in lead V3, averaging the values across three consecutive beats (Fig. 2). The QT interval was measured from the onset of the QRS complex to the end of the T wave. The R-R interval was calculated and used to compute heart rate (HR) and to correct QT distance (QTc) with Bazett’s formula [QTc = QT/(R-R)1/2] expressed in milliseconds. TpTe interval was measured from the peak of the T wave to the end of the T wave, and Bazett’s Formula was used to correct it with heart rate for TpTec.

Echocardiography examination

Echocardiographic imaging was made in the left side decubitus position in the parasternal and apical views. Two-dimensional, pulsed and color flow Doppler echocardiographic studies were conducted on all subjects by two highly trained ultrasonography technicians using the machine (CX50, Philips Ultrasound System, Andover, MA, USA). Measurements of left ventricle dimensions and volumes were performed by a computerized analysis software system. The ejection fraction (EF) was determined by the volumetric data. Analysis of the mitral inflow at the level of the tips was conducted to assess the peak early diastolic velocity (E), the late diastolic velocity (A), and the E/A ratio. The early diastolic velocity of the mitral annulus (e’) was measured at both the septal and lateral aspects, and the E/e’ ratio was subsequently calculated using the mean values of the septal E/e’ ratio and the lateral E/e’ ratio.

Statistical analyses

We retrospectively analyzed the research data, which were uploaded to a computer and evaluated via IBM SPSS v27 (SPSS Inc., Chicago, IL, USA) and MedCalc v23 (MedCalc Inc., Ostend, Belgium). Variables were presented as mean ± standard deviation. Differences in measurements between different groups with normal distribution were tested using a paired t-test or independent-samples t-test, while the data that did not fit a normal distribution were analyzed by the Wilcoxon rank sum test or Mann-Whitney U test. The mean absolute error (MAE) and mean absolute percentage error (MAPE) were employed to evaluate the accuracy of the estimation. Furthermore, the Bland-Altman plot was used to investigate the level of agreement with the 95% limits of agreement. Relationships between parameters of VR and potential factors were analyzed by linear regression and trend X2 test analysis. A two-sided P < 0.05 was considered statistically significant. Statistical power calculations were performed using the PASS software, version 15 (NCSS, LLC, Kaysville, UT, USA). The results suggested that 152 subjects would provide more than 85% power to detect differences in parameters between subgroups using a two-sided alpha of 0.05.

Results

Demographics and cardiac function

The demographic characteristics and echocardiographic data among the subjects are shown in Table 1. After exposure to a cold environment, SpO2 was decreased slightly but statistically significant (97.39 ± 1.45 vs. 96.95 ± 1.22, P = 0.004), and BP was significantly increased. Stroke volume (SV), end-diastolic volume (EDV), and end-systolic volume (ESV) were all reduced, but EF (60.63 ± 3.30 vs. 61.44 ± 3.61, P = 0.037) and cardiac output (4.43 ± 0.79 vs. 4.84 ± 0.92, P < 0.001) increased obviously. Moreover, there was no significant change in body core temperature and left ventricular diastolic function after cold exposure.

Table 1 Demographic characteristics, echocardiographic and biochemical data.
Full size table

VR changes in ECG

The changes in HR and the parameters related to VR on the ECG after cold exposure are listed in Table 2. Consistent with previous studies, HR was significantly accelerated (65.67 ± 7.37 vs. 78.47 ± 8.51, P < 0.001). And there was no significant change in QRS duration. The important ECG indicators representing VR, such as QT, QTc, TpTe, TpTec, and TpTe/QT showed different changes. QT interval was shorter (387.16 ± 24.92 vs. 365.68 ± 26.54, P < 0.001). In contrast, QTc (403.90 ± 28.47 vs. 417.66 ± 38.32, P < 0.001), TpTe interval (98.17 ± 9.26 vs. 102.29 ± 11.15, P < 0.001), and TpTec (102.36 ± 9.48 vs. 116.84 ± 14.36, P < 0.001) were greatly prolonged after exposure to a cold environment. Furthermore, the ratio of TpTe and QT was also slightly but statistically increased after cold exposure (0.25 ± 0.03 vs. 0.28 ± 0.03, P < 0.001) (Fig. 3; Table 2). The results of QTc, TpTec, and TpTe/QT measured by conventional ECG before and after cold exposure are also shown in Table 2.

Table 2 Variables of VR in ECG of the subjects before and after cold exposure.
Full size table
Fig. 3
figure 3

Changes of VR after cold exposure. VR in ECG experienced significant changes after exposure to cold environment (A). Diagrams of the changes of QTc (B), TpTec (C) and TpTe/QT (D).

Full size image

Accuracy analyses of VR

MAPE was calculated both before (HR, 1.230; QRS, 1.234; QT, 0.580; TpTe, 1.079) and after (HR, 2.124; QRS, 2.089; QT, 0.916; TpTe, 1.796) cold exposure, respectively, indicating a low average deviation between the conventional and wearable ECG (Table 3). Furthermore, a strong correlation was found between parameters measured by the two ECGs before (HR: R = 0.990, P < 0.001; QRS: R = 0.991, P < 0.001; QT: R = 0.995, P < 0.001; TpTe: R = 0.992, P < 0.001) and after (HR: R = 0.983, P < 0.001; QRS: R = 0.985, P < 0.001; QT: R = 0.995, P < 0.001; TpTe: R = 0.983, P < 0.001) cold exposure, respectively (Table 3). The results of the intraclass correlation coefficient (ICC) revealed that the parameters measured by wearable ECG had good levels of agreement with those measured by conventional ECG before (HR: ICC = 0.989, P < 0.001; QRS: ICC = 0.991, P < 0.001; QT: ICC = 0.996, P < 0.001; TpTe: ICC = 0.993, P < 0.001) and after (HR: ICC = 0.984, P < 0.001; QRS: ICC = 0.985, P < 0.001; QT: ICC = 0.996, P < 0.001; TpTe: ICC = 0.987, P < 0.001) cold exposure, respectively (Table 3). Additionally, the Bland-Altman plots demonstrated a small bias of wearable ECG compared to conventional ECG before and after cold exposure, respectively (Fig. 4).

Table 3 The correlations and differences of VR between conventional and wearable ECG.
Full size table
Fig. 4
figure 4

Bland–Altman plots of variables for VR between measured by conventional ECG and wearable ECG. HR, QRS, QT and TpTe before cold exposure respectively (A–D) ; HR, QRS, QT and TpTe after cold exposure respectively (E–H). Mean biases (solid line) and 95% limits of agreement (dashed line) were also depicted.

Full size image

Predictors of VR changes

To identify predictors of VR changes, both univariate and multivariate linear regression analyses were employed. On univariate analysis, baseline HR and LV end diastolic diameters were associated with QTc variation (Table 4); baseline HR, EDV, ESV, SV, and living history of cold zone were associated with TpTec variation (Table 4); baseline HR, EDV, and ESV were associated with TpTe/QT variation (Table 4). On multivariate analysis, those with higher baseline HR were at increased risk of prolonged QTc (β = 0.247; P = 0.002), TpTec (β = 0.462; P < 0.001), and TpTe/QT (β = 0.242; P = 0.003) when compared to those with lower baseline HR. The regression lines showing relationships between baseline HR and different VR variables are presented in Fig. 5.

Table 4 Results of univariate and multivariate linear regression analyses for the variations of qtc, TpTec and tpte/qt.
Full size table
Fig. 5
figure 5

The linear regression plots between the variations of VR and baseline HR. Correlation between baseline HR and ΔQTc (A), ΔTpTec (B) and ΔTpTe/QT (C), respectively. Δ means value for variation after cold exposure. The coefficient of determination (R) and 95% confidence interval bounds (dotted line) were also depicted.

Full size image

Trend test of baseline HR and VR changes

The effects of different baseline HR on VR variations are presented in Fig. 6. Subjects were grouped according to different baseline HR, and as baseline HR increased, QTc (P < 0.001), TpTec (P < 0.001), and TpTe/QT (P < 0.001) variation all showed increasing trends, which were statistically significant.

Fig. 6
figure 6

The effects of different baseline HR on the variations of VR. Trend X2 test for ΔQTc (A), ΔTpTec (B) and ΔTpTe/QT (C), respectively. Δ means value for variation after cold exposure. P* means P value for trend.

Full size image

Discussion

In the present study, a self-developed 12-lead handheld wearable ECG was employed, demonstrating that this device was a feasible and accurate device for electrocardiographic assessments in an extreme cold environment. Furthermore, we conducted a comparative analysis of the VR indexes before and after cold exposure. Additionally, this is the first study to reveal the predictive effect of baseline heart rate on repolarization during cold exposure. When compared to those with lower heart rates, When compared to those with lower heart rates, elevated baseline HR independently predicted prolonged QTc and TpTec, suggesting a mechanistic link between sympathetic overactivity and repolarization instability under cold stress.

With the advancement of polar development strategies by countries worldwide, the number of people entering and stationed in extreme cold regions has increased annually. It has been indicated that prolonged or excessive cold stress could activate the sympathetic nervous system19. In general, sympathetic activation may lead to electrical abnormalities in the heart and elevate the risk of sudden cardiac death20. Therefore, conducting risk assessments for these populations appears to be meaningful. VR, as a complex electrical phenomenon which represents a crucial stage in electrical cardiac activity, is expressed on the surface ECG by the interval between the start of the QRS complex and the end of the T wave or U wave (QT)21. Research has demonstrated that small perturbations in this process can potentially trigger malignant arrhythmias, which may result in sudden cardiac death. These arrhythmias can occur in apparently healthy individuals or may be associated with underlying conditions22. Therefore, the analysis of VR seems to represent an interesting tool to implement risk stratification of arrhythmic events in different settings. In recent decades, some VR markers such as QT, QTc, TpTe, TpTec and TpTe/QT obtained from 12-lead surface ECG, have been found to be useful to predict malignant cardiac arrhythmias in several clinical conditions11,12. However, few studies on VR in the real-world outside cold environments have been reported. This This highlights the necessity for further research in this field.

Previous experimental or clinical studies have examined the effects of cold on cardiac electrical function. These studies revealed various impacts on the ECG, including ventricular arrhythmias, interval prolongation, T-wave abnormalities, and significant ST-segment depression induced by cold23,24which varied with the form, duration, and intensity of the cold stress. Hintsala et al. reported that short-term cold exposure leads to an elongation of cardiac repolarization, as evidenced by the prolongation of the TpTe interval, without significantly affecting the QT interval25. In addition to the differences in the population included, the cold exposure sites in this study involved superficial facial areas, as opposed to our study (via the respiratory tract and the forehead), which would lead to the co-activation of the autonomic nervous system’s sympathovagal components (parasympathetic and sympathetic), and therefore a certain difference from our conclusions. Furthermore, we concentrated on changes in VR under extremely cold conditions. Consequently, compared to the previous study, our environmental condition was at a lower temperature (-35℃ vs. -10℃) and the exposure lasted longer (2 h vs. 15 min). Our aim was to provide additional clinical evidence on the relationship between cold stress and ECG changes. Besides, ECG parameters are typically rate dependent. The QT interval is recognized to be highly dependent on HR, and its rate dependency can vary with changes in autonomic function. In contrast, TpTe has been shown to have a minor or negligible connection with HR26. Thus, we conducted HR correction for QT interval and TpTe interval and found that QTc and TpTec interval were obviously increased after cold exposure. A recent study involving the cold pressor test also found a notable increase in QTc after cold stimulation, which aligns with our findings27.

Recently, wearable ECG devices have undergone enhancements and have become more widely adopted, evolving from single-lead to 12-lead systems16. These devices have been demonstrated to possess significant clinical application value and promising prospects in the management of cardiovascular diseases28. To date, commercial wearable devices utilize two primary technologies for heart rate and rhythm monitoring. One approach employs the photoplethysmography (PPG) technique, enabling the detection of these health metrics via a wristband or smartwatch, including the Apple Watch Series 4 (manufactured by Apple Inc.), which has obtained approval from the US Food and Drug Administration for identifying atrial fibrillation29. Furthermore, several studies have demonstrated a high level of concordance in parameters, including the ST-segment and QT interval, between the smartwatch ECG and the standard ECG29,30,31,32,33. Nonetheless, PPG technology is influenced by a multitude of factors34. Cold stress leads to the constriction of blood vessels on the body’s surface, resulting in slower blood circulation and changes in the absorption and reflection of light. Consequently, the light signal received by the PPG sensor weakens, which affects the accuracy of detection. Although the information from a single lead is still limited, some studies have also attempted to obtain a multi-lead ECG by repositioning the devices on the body, akin to obtaining a standard ECG30,32,35. Another approach is the patch-style ECG device based on the Mason-Likar leads, which is more portable and less disruptive to normal life compared to conventional ECG. Based on large-scale datasets and combined with machine learning, the diagnostic results of wearable multi-ECG devices demonstrated great accuracy, fully reflecting the dynamics of the heart36,37,38. This study evaluated a 12-lead wearable ECG with a simple wear process. We utilized a fabric electrode in conjunction with a novel “sandwich” design in our self-developed device to optimize performance in cold environments. The verification of accuracy suggested that it held substantial application value in outdoor extreme cold environments. The device is expected to offer advantages in scenarios such as disaster relief, and is crucial for identifying early ECG abnormalities among the working population in extreme environments.

Few previous studies have reported on the risk factors of cardiac repolarization after exposure to a cold environment. In our study, we demonstrated that those in the higher baseline HR were at increased risk of prolonged VR after cold exposure when compared to those with lower HR. This may be due to a high baseline HR associated with exaggerated energy expenditure, impaired myocardial oxygen delivery due to a short diastole period, and loss of the positive force-frequency relationship (Bowditch effect)39. A higher level of baseline HR indicates a dominant sympathetic system state, which may result in increased norepinephrine secretion by the sympathetic nervous system. This excessive increase in cardiac afterload and decompensation of cardiac function can lead to a mismatch between oxygen supply and demand. Our previous research also found that subjects with heightened sympathetic activity (higher BPV levels) were more likely to develop cardiovascular abnormalities during exposure to an extremely cold environment40. Additionally, in other extreme environments, such as high altitude with hypobaric and hypoxic conditions, those with a low baseline HR adapted well by effectively utilizing HR reserve and enhancing left ventricular contractility41. Interestingly, our results also showed a minor but still significant decrease in SpO2 after a period of cold exposure, which may be due to decreased blood circulation and respiratory function caused by cold stress. In the early time, E A Koller et al. found that the lengthening of QT is probably due to the direct effects of hypoxia42. A temporary and rapid decrease in oxygen supply to cardiac myocytes was not energy limiting and does not deplete ATP, but can alter the function of several cardiac ion channels43. These ion channels have been proposed to affect the process of VR. In sum, the changes of myocardial oxygen consumption, biological functions of cardiomyocytes, myocardial cell membrane ion channels, and sympathovagal interactions may be the potential reasons for the prolongation of cardiac repolarization in a cold environment.

Limitations

This study has several limitations that warrant consideration. First, our use of Bazett’s formula for QT interval correction assumes a constant inverse relationship between heart rate and QT interval, which may not always hold true in all physiological states. Second, the absence of longitudinal follow-up precludes assessment of whether VR prolongation translates to actual arrhythmic events, a critical gap given the observational nature of the study. Therefore, we are unable to evaluate the clinical implications of our findings, i.e. we do not know if the participants are at an increased risk of arrhythmia or sudden cardiac death even though it is well established that the prolongation of VR markers are related to the high risk of malignant arrhythmia in previous studies and guidelines. Third, QT and TpTe interval measurements were calculated manually instead of computer-assisted calculations, but several studies have suggested that manual measurement on paper-printed ECGs obtained at a standard signal size and paper speed may have certain degrees of accuracy and repeatability problems. Lastly, the present results still cannot be extended to the entire population. In the long term, conducting long-term validation research within a large and diverse population is the scope of future study, as wearable devices will be extensively used by the public.

Conclusions

In this study, we independently developed a 12-lead wearable ECG. To optimize the application of ECG in cold environments, we utilized the fabric electrode combined with a novel “sandwich” design. The wearable ECG demonstrated good accuracy and agreement with conventional ECG, making it highly significant for identifying early ECG abnormalities in individuals working in an outdoor extreme cold environment. Moreover, we have documented the effects of VR changes following extreme cold exposure in a real-world setting, with a considerable sample size. Our findings indicated that cold stress could lead to an increase in the QTc and TpTec intervals, as well as in the ratio of TpTe/QT. We suggest that a lower HR offers a protective effect against prolonged VR. This dual-focused study not only validates a wearable ECG for extreme cold environments but also identifies baseline heart rate as a modifiable risk factor for repolarization instability, offering actionable strategies (e.g., pre-exposure HR control) for populations in polar or high-altitude settings.