Next Article in Journal
Association of Lifestyle Behaviors with Quality of Life in Patients with COPD: A Cross-Sectional Study in Primary Care
Previous Article in Journal
Reconstructive Surgery of the Head and Neck in Organ Transplant Recipients: A Case Report and a Review of the Literature
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
JCM
Volume 13
Issue 16
10.3390/jcm13164792
jcm-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Screening for RV Dysfunction Using Smartphone ECG Analysis App: Validation Study with Acute Pulmonary Embolism Patients

by
Yoo Jin Choi
1,†,
Min Ji Park
1,†,
Youngjin Cho
2,3,
Joonghee Kim
3,4,
Eunkyoung Lee
3,
Dahyeon Son
3,
Seo-Yoon Kim
1 and
Moon Seung Soh
5,*
1
Department of Emergency Medicine, Ajou University School of Medicine, 164, World cup-ro, Yeongtong-gu, Suwon-si 16499, Gyeonggi-do, Republic of Korea
2
Cardiovascular Center, Department of Internal Medicine, Seoul National University Bundang Hospital, 82, Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 13620, Gyeonggi-do, Republic of Korea
3
ARPI Inc., Room 12 Startup Incubation Center, 172, Dolma-ro, Bundang-gu, Seongnam-si 13605, Gyeonggi-do, Republic of Korea
4
Department of Emergency Medicine, Seoul National University Bundang Hospital, 82, 166 Gumi-ro 173 Beon-gil, Bundang-gu, Seongnam-si 13620, Gyeonggi-do, Republic of Korea
5
Department of Cardiology, Ajou University School of Medicine, 164, World cup-ro, Yeongtong-gu, Suwon-si 16499, Gyeonggi-do, Republic of Korea
*
Author to whom correspondence should be addressed.
†
These authors contributed equally to this work.
J. Clin. Med. 2024, 13(16), 4792; https://doi.org/10.3390/jcm13164792
Submission received: 11 July 2024 / Revised: 9 August 2024 / Accepted: 13 August 2024 / Published: 14 August 2024
(This article belongs to the Section Emergency Medicine)
Browse Figures
Versions Notes

Abstract

:
Background: Acute pulmonary embolism (PE) is a critical condition where the timely and accurate assessment of right ventricular (RV) dysfunction is important for patient management. Given the limited availability of echocardiography in emergency departments (EDs), an artificial intelligence (AI) application that can identify RV dysfunction from electrocardiograms (ECGs) could improve the treatment of acute PE. Methods: This retrospective study analyzed adult acute PE patients in an ED from January 2021 to December 2023. We evaluated a smartphone application which analyzes printed ECGs to generate digital biomarkers for various conditions, including RV dysfunction (QCG-RVDys). The biomarker’s performance was compared with that of cardiologists and emergency physicians. Results: Among 116 included patients, 35 (30.2%) were diagnosed with RV dysfunction. The QCG-RVDys score demonstrated significant effectiveness in identifying RV dysfunction, with a receiver operating characteristic–area under the curve (AUC) of 0.895 (95% CI, 0.829–0.960), surpassing traditional biomarkers such as Troponin I (AUC: 0.692, 95% CI: 0.536–0.847) and ProBNP (AUC: 0.655, 95% CI: 0.532–0.778). Binarized based on the Youden Index, QCG-RVDys achieved an AUC of 0.845 (95% CI: 0.778–0.911), with a sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of 91.2% (95% CI: 82.4–100%), 77.8% (95% CI: 69.1–86.4%), 63.3% (95% CI: 54.4–73.9%), and 95.5% (95% CI: 90.8–100%), respectively, significantly outperforming all the expert clinicians, with their AUCs ranging from 0.628 to 0.683. Conclusions: The application demonstrates promise in rapidly assessing RV dysfunction in acute PE patients. Its high NPV could streamline patient management, potentially reducing the reliance on echocardiography in emergency settings.

1. Introduction

Acute pulmonary thromboembolism (PE) is a serious emergency condition that can be life-threatening. Screening right ventricular (RV) dysfunction is crucial in the management of acute PE [1], as it significantly impacts patient outcomes. The RV’s response to increased afterload can result in myocardial ischemia and heart failure, increasing the risk of hemodynamic collapse and mortality [2]. Therefore, the timely and accurate detection of RV dysfunction is vital for risk stratification and for guiding treatment in acute PE.
Echocardiography is the standard tool for evaluating RV function, providing detailed hemodynamic information [3,4,5]. However, the fast-paced environment of the emergency department (ED) requires faster and simpler diagnostic methods, and in that regard, echocardiography may be limited in its use in the ED due to its reliance on skilled operators and the availability of equipment.
One potential solution to address this issue is to utilize electrocardiograms (ECGs). Recognizing RV strain patterns such as new right-axis deviations, the S1Q3T3 pattern, or ST-segment depressions, with T-wave inversions in leads V1 to V3 and leads II, III, and aVF from ECGs, can facilitate the evaluation of RV dysfunction [6,7]. However, these methods have limitations in accuracy, reducing their utility.
The integration of digital technology into medicine, and especially the use of artificial intelligence (AI) in acute care, offers new possibilities. AI solutions that analyze ECG data represent a shift toward more accessible and rapid cardiac assessments [8,9]. However, these applications typically require raw digital ECG data, which is impractical in real-world clinical settings where only printed ECG data are available.
To bridge this gap, we developed ECG BuddyTM, a mobile application that generates ten digital biomarkers by analyzing images of printed ECGs. Previous studies have suggested this tool’s utility in various emergency situations, including suspected myocardial infarctions and severe hyperkalemia [10,11,12,13]. This study aims to evaluate the application’s capability in identifying RV dysfunction in ED patients with acute PE. Additionally, we will compare its performance to those of expert clinicians.

2. Methods

2.1. Study Design and Data Collection

This study analyzed adult patients (≥18 years) with acute PE presenting at the ED of a tertiary hospital (Ajou University Medical Center) from January 2021 to December 2023 to validate AI software (version number 1.0.4.X) initially developed using data from a different hospital (Seoul National University Bundang Hospital). Patients were excluded if they did not have an ECG within 72 h post-ED arrival or an echocardiogram from 72 h before to 72 h after the ECG. Because one of the standard diagnostic tests for PE is a contrast-enhanced chest computed tomography (CT) scan, which is also commonly used in research facilities, the patients included in this study had their PE diagnosed by contrast-enhanced chest CT scan [2]. Two board-certified emergency physicians manually reviewed electronic medical records (EMR) to determine eligibility and gather information on demographics, PE risk factors, vital signs, lab results, and heart rhythm. The ECG data were collected from the EMR by manually cropping the waveform areas of each ECG report to remove any identifying information. The institutional review board approved this study (IRB No.: AJOUIRB-DB-2024-177), waiving informed consent due to its retrospective design.

2.2. Assessment of RV Function

The presence of RV dysfunction was determined by a qualitative review of echocardiographic reports considering the presence of a D-shaped left ventricle, Mcconnell’s sign, or decreased fractional area change (FAC < 35%). The FAC was calculated as (end-diastolic area–end-systolic area)/end-diastolic area x 100 in the RV-focused view of each echocardiogram [5]. In addition, measurements of echocardiographic right ventricular systolic pressure (RVSP), an indicator of pulmonary hypertension, were collected and categorized into four groups: RVSP I (less than 35 mmHg), RVSP II (35–49 mmHg), RVSP III (50–64 mmHg), and RVSP IV (greater than 64 mmHg). RVSP was determined from the peak tricuspid regurgitant jet velocity (TR Vpeak), using the simplified Bernoulli equation and combining this value with an estimate of the right atrium (RA) pressure estimated from the inferior vena cava diameter and the respiratory changes in the subcostal view [RVSP = 4(TR Vpeak)2 + RA pressure].

2.3. ECG Analysis by AI Application

An AI smartphone application, named “ECG Buddy”, was used to analyze the ECG images. The application, approved by the Korean MFDS and freely available for download in Korean appstores, can analyze 12-lead ECGs by taking pictures of ECG images to produce 10 digital biomarkers, (Quantitatve ECG[QCG®] scores ranging from 0 to 100) for various emergencies and cardiac dysfunctions [14]. We analyzed the ECG capture images by first displaying them on a desktop monitor and taking pictures of the ECG images using the application. We recorded digital biomarkers for RV dysfunction (QCG-RVDys) and pulmonary hypertension (QCG-PHTN) for each ECG for evaluation.

2.4. Expert Analysis of ECGs

To establish a benchmark for the AI biomarkers, expert evaluations of the same ECGs were obtained from a group of two cardiologists and three emergency physicians, all blinded to the patients’ clinical information. The experts, all board-certified physicians with at least 8 years of clinical experiences, asked to review each ECG image freely without a time limit to determine whether the ECG exhibited RV dysfunction primarily using an S1Q3T3 pattern [15,16].

2.5. Statistical Analysis

The primary metric used to assess the performance of the biomarkers was receiver operating characteristic–area under the curve (ROC-AUC). We compared the ROC-AUC of QCG-RVDys for identifying RV dysfunction to those for Troponin I and proBNP using original measurements in a continuous scale. For comparison with the experts, we binarized the biomarker using the threshold that maximized its Youden index (binarized QCG-RVDys), as the experts were asked to give their opinion in a binary format (yes or no). The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of the biomarker were calculated. The ROC-AUC of QCG-PHTN in identifying moderate pulmonary hypertension, as determined by RVSP 50 mmHg or more, was calculated too; however, the performance of the biomarker was not compared to other measurements, as it was not the main study topic of this study, and there is no popular method for identifying the condition clinically. All data analyses were conducted using R software, version 4.1.0.

3. Results

From January 2021 to December 2023, 131 patients with acute PE were admitted to the emergency room. After the exclusion of fifteen patients lacking ECGs or echocardiography measurements meeting the eligibility criteria, a total of 116 patients were included in this study (Figure 1). Within the patient population, 35 patients were assessed as having RV dysfunction (RVD group, 30.2%) and 81 patients were assessed as not having RV dysfunction (No RVD group, 69.8%).
There were no statistically significant differences between the two groups in terms of previous risk factors, except for the number of patients with a history of deep vein thrombosis (No RVD: 0, RVD: 3, p = 0.042) (Table 1). Initial vital signs measured in the ED showed no significant differences. However, in blood tests, aspartate transaminase (AST; No RVD: 22.5 U/L vs. RVD: 32.0 U/L, p = 0.003), alanine transaminase (ALT; No RVD: 19.0 U/L vs. RVD: 30.0 U/L, p = 0.017), Troponin I (No RVD: 0.052 μg/mL vs. RVD: 0.176 μg/mL, p = 0.034), and proBNP (No RVD: 482.0 pg/mL vs. RVD: 1366.0 pg/mL, p = 0.024) were significantly higher in the RVD group. In addition, the RVSP group distribution was also significantly different (p < 0.001), and there was a trend of echocardiography being performed earlier in the RVD group (ED arrival to echocardiography, No RVD: 22.1 h vs. RVD: 16.1 h, p= 0.036).
The QCG-RVDys scores were significantly different between the No RVD group (6.8 [2.5–22.5]) and the RVD group (78.7 [35.7–94.9]), with a p-value of <0.001 (Figure 2, Table 2). The QCG-PHTN scores also showed significant differences across the RVSP groups I, II, III, and IV, demonstrating a clear correlation with increasing RVSP (p < 0.001, p-trend < 0.001, Table 2 and Figure S1).
The AUC-ROC for QCG-RVDys in diagnosing RV dysfunction was 0.895 (0.829–0.960), significantly higher than that for Troponin I, 0.692 (0.536–0.847), and ProBNP, 0.655 (0.532–0.778) (p = 0.046 and p = 0.001, respectively, Figure 3, Table 3). When the QCG-RVDys was binarized at a threshold of 24.65, based on the Youden index, the AUC was 0.845 (0.778–0.911), with a sensitivity, specificity, PPV, and NPV of 91.2% (82.4–100), 77.8% (69.1–86.4), 63.3% (54.4–73.9), and 95.5% (90.8–100), respectively. In comparison, the five experts showed AUCs ranging from 0.628 to 0.683, all statistically significantly lower than that of the binarized QCG-RVD. (Table S1).
The performance of QCG-PHTN in predicting elevated RVSP (RVSP groups III and IV; RVSP ≥ 50 mmHg) showed an ROC-AUC of 0.820 (0.728–0.912) (Table S2, Figure S2). At a threshold of 0.2590, the sensitivity, specificity, PPV, and NPV were 82.6% (65.2–95.7), 72.8% (64.1–81.5), 43.2% (34.7–54.1), and 94.4% (89.4–98.6), respectively.
To explore the potential factors influencing the QCG-RVDys score, we divided the patient population into two groups based on a threshold score of 24.65. Table S3 compares the characteristics of patients with QCG-RVDys scores above and below this threshold. We observed that patients with elevated QCG-RVDys scores (>24.65) were more likely to have higher levels of Troponin I and ProBNP, as well as higher RVSP in echocardiography, compared to those with lower scores. Additionally, differences in heart rhythm and laboratory measurements, such as white blood cell count and lactate levels, were noted between the two groups.

4. Discussion

This study showed that the digital ECG biomarker produced by a smartphone application can predict RV dysfunction with high accuracy in patients with acute PE using only printed ECGs without additional clinical information. We have also shown it can also predict pulmonary hypertension, too, expanding the utility of ECGs. This is the first study to assess critical cardiac functions in patients with acute PE using ECG AI, especially through a smartphone app, distinguishing it from similar research in the field.
Based on the results, smartphone-based ECG analysis software (version number 1.0.4.X) could be considered for use in treatment decisions for patients with acute PE in EDs. Assessing the right heart function, especially RV dysfunction and pulmonary hypertension, is crucial for determining the treatment plan for acute PE, alongside evaluating hemodynamic stability. This can rapidly guide decisions regarding thrombolytic therapy or surgical embolectomy in patients with RV dysfunction and hemodynamic instability [2,17,18]. According to our study, QCG-RVDys has a high NPV of 95.5 (90.8–100%), suggesting it could be useful in excluding the possibility of RV dysfunction in patients with newly diagnosed acute PE, for whom an immediate echocardiography is not feasible.
This utility can also be extended to the early diagnosis of massive PE. Patients with massive PE often present with nonspecific symptoms like dizziness, syncope, dyspnea, and chest pain, necessitating a broad differential diagnosis. In many EDs, when patients present with such symptoms, ECG is one of the first triage tests usually performed. Utilizing this, the early detection of RV dysfunction, often associated with massive PE, could prioritize the suspicion of PE, allowing for early diagnostic imaging, such as contrast chest CT, and hastening diagnosis [2]. Recently, studies have demonstrated the growing potential of AI models trained specifically to identify PE using ECG data [19,20,21,22,23,24]. In contrast, our model is designed to identify RV dysfunction, a critical characteristic of massive PE, which could serve as an early indicator in the diagnostic process. However, further studies are needed to determine which approach—the direct identification of PE or the detection of associated RV dysfunction—is more effective for early diagnosis and for improving patient outcomes.
Similarly, the evaluation of RV dysfunction could be expanded to other diseases, such as in acute respiratory distress syndrome and or pulmonale, where the assessment of RV strain is beneficial for establishing lung-protective ventilation strategies, balanced fluid therapy, the selection of vasopressors (minimizing impact on pulmonary vascular resistance), pulmonary vasodilation (e.g., NO, sildenafil), and effective monitoring strategies (echocardiography or hemodynamic measurement) [25,26,27,28,29]. Likewise, the early recognition of RV dysfunction in RV myocardial infarction through AI could aid safer decision making regarding early fluid therapy and vasodilator use. However, the efficacy and safety of the application in these specific clinical scenarios have not been evaluated, indicating a need for further research.
In the RVD group, the proportion of patients with RVSP >50 mmHg was higher, and the QCG-PHTN score increased with increasing RVSP, showing good utility. However, there was a wider distribution of QCG-PHTN scores in the high-RVSP group (especially >64 mmHg) compared to the relatively low-RVSP group, which can be explained by the fact that RVSP is not an absolute reflection of pulmonary hypertension and is influenced by the hemodynamics of the RV. In echocardiography, RVSP is mainly calculated from the tricuspid regurgitation maximal velocity (TR Vmax) and the collapsibility of the inferior vena cava [5]. In patients with RV dysfunction, the TR Vmax is also reduced, because the contractile force of the RV is reduced, which may cause this result.
The observed lower PPV of the QCG-RVDys score may be influenced by several patient characteristics that differ between those with scores above and below the threshold of 24.65. While higher Troponin I, ProBNP levels, and RVSP in the high-score group are consistent with the clinical features of PE, the increased prevalence of tachycardia and arrhythmias in this group could potentially influence the QCG-RVDys score. Although it is not yet clear whether these ECG abnormalities directly contribute to higher QCG-RVDys scores, their presence as common findings in emergency department patients may play a role. Further research is needed to clarify these potential confounding factors and refine the score’s predictive accuracy.
Handheld ultrasonography devices are increasingly being used to rapidly assess RV function at the bedside in patients with PE. These devices can detect key indicators of RV dysfunction, such as a dilated RV and D-sign, which are important in the evaluation of PE [30,31,32,33]. Integrating ECG Buddy™ with handheld ultrasonography could further improve PE diagnosis, and future studies should explore this combined approach to enhance early diagnosis and improve patient outcomes.
This study has several limitations. Firstly, it is a single-center, retrospective study with a small sample size, which may limit the generalizability of our findings. Therefore, a larger, multicenter, prospective study is needed. Second, we defined RV dysfunction based on echocardiography, which relies on the expertise of a trained examiner, and RVSP is not an absolute reflection of pulmonary hypertension, as it is dependent on RV dysfunction and cardiac physiology, and therefore needs to reflect a more appropriate endpoint. Finally, since this study was based on patients with already-diagnosed PE, a real-world point-of-care screening test would need to incorporate a variety of patient clinical conditions.

5. Conclusions

In conclusion, when predicting the RV dysfunction of digital ECG biomarkers in patients with acute PE using ECGs, smartphone software can more accurately assess the presence or absence of RV dysfunction compared to traditional methods by clinical specialists. Particularly, this smartphone software demonstrates a high negative predictive value, suggesting the potential to omit or delay costly and time-consuming echocardiography in patients of low risk.

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/jcm13164792/s1, Figure S1. QCG-PHTN score across RVSP groups (RVSP, right ventricular systolic pressure). Figure S2. Performance of QCG-PHTN on increased RVSP (RVSP ≥ 50 mmHg). Table S1. Performance of binarized QCG-RVDys and human experts on identifying RV dysfunction from ECG. Table S2. performance of binarized QCG-PHTN on increased RVSP (RVSP ≥ 50 mmHg). Table S3. Patient characteristics by RV dysfunction score.

Author Contributions

Conceptualization, Y.J.C., M.J.P., J.K. and M.S.S.; Data curation, Y.J.C., M.J.P. and S.-Y.K.; Formal analysis, Y.J.C., M.J.P. and M.S.S.; Methodology, Y.C., J.K., E.L. and D.S.; Writing—original draft, Y.J.C. and M.J.P.; Writing—review and editing, Y.C., J.K., S.-Y.K. and M.S.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research was partly supported by a grant of the Technological Innovation R&D Program (SCALEUP TIPS), funded by the Ministry of SMEs and Startups (grant number: RS-2024-00415492), and the Medical AI Clinic Program through the NIPA, funded by the MSIT (grant number: H0904-24-1002).

Institutional Review Board Statement

This study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of Ajou Medical Center (IRB No.: AJOUIRB-DB-2024-177, date of approval 15 April 2024).

Informed Consent Statement

Patient consent was waived due to the study being a retrospective observational study based on medical records and not including private personal information.

Data Availability Statement

The original contributions presented in this study are included in this article. Further inquiries can be directed to the corresponding author/s.

Conflicts of Interest

J.K. developed the algorithm and founded the start-up company ARPI Inc. He is the CEO of the company. Y.C. works for the company as a research director. E.L. and D.S. work for the company as data scientists. M.J.P. and Y.G.C. hold stocks in ARPI Inc. Otherwise, there are no conflicts of interest for the other authors.

References

  1. Members, A.F.; Torbicki, A.; Perrier, A.; Konstantinides, S.; Agnelli, G.; Galiè, N.; Pruszczyk, P.; Bengel, F.; Brady, A.J.B.; Ferreira, D.; et al. Guidelines on the Diagnosis and Management of Acute Pulmonary EmbolismThe Task Force for the Diagnosis and Management of Acute Pulmonary Embolism of the European Society of Cardiology (ESC). Eur. Heart J. 2008, 29, 2276–2315. [Google Scholar] [CrossRef]
  2. Konstantinides, S.V.; Meyer, G.; Becattini, C.; Bueno, H.; Geersing, G.-J.; Harjola, V.-P.; Huisman, M.V.; Humbert, M.; Jennings, C.S.; Jiménez, D.; et al. 2019 ESC Guidelines for the Diagnosis and Management of Acute Pulmonary Embolism Developed in Collaboration with the European Respiratory Society (ERS)The Task Force for the Diagnosis and Management of Acute Pulmonary Embolism of the European Society of Cardiology (ESC). Eur. Heart J. 2019, 41, 543–603. [Google Scholar] [CrossRef]
  3. Ryan, T.; Petrovic, O.; Dillon, J.C.; Feigenbaum, H.; Conley, M.J.; Armstrong, W.F. An Echocardiographic Index for Separation of Right Ventricular Volume and Pressure Overload. J. Am. Coll. Cardiol. 1985, 5, 918–924. [Google Scholar] [CrossRef] [PubMed]
  4. McConnell, M.V.; Solomon, S.D.; Rayan, M.E.; Come, P.C.; Goldhaber, S.Z.; Lee, R.T. Regional Right Ventricular Dysfunction Detected by Echocardiography in Acute Pulmonary Embolism. Am. J. Cardiol. 1996, 78, 469–473. [Google Scholar] [CrossRef] [PubMed]
  5. Rudski, L.G.; Lai, W.W.; Afilalo, J.; Hua, L.; Handschumacher, M.D.; Chandrasekaran, K.; Solomon, S.D.; Louie, E.K.; Schiller, N.B. Guidelines for the Echocardiographic Assessment of the Right Heart in Adults: A Report from the American Society of Echocardiography Endorsed by the European Association of Echocardiography, a Registered Branch of the European Society of Cardiology, and the Canadian Society of Echocardiography. J. Am. Soc. Echocardiogr. 2010, 23, 685–713. [Google Scholar] [CrossRef] [PubMed]
  6. Rodger, M.; Makropoulos, D.; Turek, M.; Quevillon, J.; Raymond, F.; Rasuli, P.; Wells, P.S. Diagnostic Value of the Electrocardiogram in Suspected Pulmonary Embolism. Am. J. Cardiol. 2000, 86, 807–809. [Google Scholar] [CrossRef] [PubMed]
  7. Geibel, A.; Zehender, M.; Kasper, W.; Olschewski, M.; Klima, C.; Konstantinides, S.V. Prognostic Value of the ECG on Admission in Patients with Acute Major Pulmonary Embolism. Eur. Respir. J. 2005, 25, 843–848. [Google Scholar] [CrossRef] [PubMed]
  8. Yao, X.; Rushlow, D.R.; Inselman, J.W.; McCoy, R.G.; Thacher, T.D.; Behnken, E.M.; Bernard, M.E.; Rosas, S.L.; Akfaly, A.; Misra, A.; et al. Artificial Intelligence–Enabled Electrocardiograms for Identification of Patients with Low Ejection Fraction: A Pragmatic, Randomized Clinical Trial. Nat. Med. 2021, 27, 815–819. [Google Scholar] [CrossRef] [PubMed]
  9. Attia, Z.I.; Kapa, S.; Yao, X.; Lopez-Jimenez, F.; Mohan, T.L.; Pellikka, P.A.; Carter, R.E.; Shah, N.D.; Friedman, P.A.; Noseworthy, P.A. Prospective Validation of a Deep Learning Electrocardiogram Algorithm for the Detection of Left Ventricular Systolic Dysfunction. J. Cardiovasc. Electrophysiol. 2019, 30, 668–674. [Google Scholar] [CrossRef]
  10. Choi, Y.J.; Park, M.J.; Ko, Y.; Soh, M.-S.; Kim, H.M.; Kim, C.H.; Lee, E.; Kim, J. Artificial Intelligence versus Physicians on Interpretation of Printed ECG Images: Diagnostic Performance of ST-Elevation Myocardial Infarction on Electrocardiography. Int. J. Cardiol. 2022, 363, 6–10. [Google Scholar] [CrossRef]
  11. Kim, D.; Jeong, J.; Kim, J.; Cho, Y.; Park, I.; Lee, S.-M.; Oh, Y.T.; Baek, S.; Kang, D.; Lee, E.; et al. Hyperkalemia Detection in Emergency Departments Using Initial ECGs: A Smartphone AI ECG Analyzer vs. Board-Certified Physicians. J. Korean Méd. Sci. 2023, 38, e322. [Google Scholar] [CrossRef]
  12. Lee, S.H.; Hong, W.P.; Kim, J.; Cho, Y.; Lee, E. Smartphone AI vs. Medical Experts: A Comparative Study in Prehospital STEMI Diagnosis. Yonsei Méd. J. 2024, 65, 174–180. [Google Scholar] [CrossRef]
  13. Park, M.J.; Choi, Y.J.; Shim, M.; Cho, Y.; Park, J.; Choi, J.; Kim, J.; Lee, E.; Kim, S.-Y. Performance of ECG-Derived Digital Biomarker for Screening Coronary Occlusion in Resuscitated Out-of-Hospital Cardiac Arrest Patients: A Comparative Study between Artificial Intelligence and a Group of Experts. J. Clin. Med. 2024, 13, 1354. [Google Scholar] [CrossRef] [PubMed]
  14. Ministry of Food and Drug Safety. Available online: https://emedi.mfds.go.kr/search/data/MNU20237 (accessed on 5 April 2024).
  15. Barman, H.A.; Atici, A.; Sahin, I.; Dogan, O.; Okur, O.; Tugrul, S.; Avci, I.; Yildirmak, M.T.; Gungor, B.; Dogan, S.M. Prognostic Value of Right Ventricular Strain Pattern on ECG in COVID-19 Patients. Am. J. Emerg. Med. 2021, 49, 1–5. [Google Scholar] [CrossRef]
  16. Raad, M.; Gorgis, S.; Dabbagh, M.; Chehab, O.; Parikh, S.; Singh, G. Right Heart Strain on Presenting 12-Lead Electrocardiogram Predicts Critical Illness in COVID-19. JACC Clin. Electrophysiol. 2021, 7, 485–493. [Google Scholar] [CrossRef] [PubMed]
  17. Jaff, M.R.; McMurtry, M.S.; Archer, S.L.; Cushman, M.; Goldenberg, N.; Goldhaber, S.Z.; Jenkins, J.S.; Kline, J.A.; Michaels, A.D.; Thistlethwaite, P.; et al. Management of Massive and Submassive Pulmonary Embolism, Iliofemoral Deep Vein Thrombosis, and Chronic Thromboembolic Pulmonary Hypertension. Circulation 2011, 123, 1788–1830. [Google Scholar] [CrossRef] [PubMed]
  18. Kucher, N.; Goldhaber, S.Z. Management of Massive Pulmonary Embolism. Circulation 2005, 112, e28–e32. [Google Scholar] [CrossRef] [PubMed]
  19. Chen, Y.-C.; Tsai, S.-C.; Lin, C.; Lin, C.-S.; Fang, W.-H.; Lou, Y.-S.; Lee, C.-C.; Liu, P.-Y. A Deep-Learning Algorithm-Enhanced Electrocardiogram Interpretation for Detecting Pulmonary Embolism. Acta Cardiol. Sin. 2022, 39, 913–928. [Google Scholar] [CrossRef]
  20. Kalmady, S.V.; Salimi, A.; Sun, W.; Sepehrvand, N.; Nademi, Y.; Bainey, K.; Ezekowitz, J.; Hindle, A.; McAlister, F.; Greiner, R.; et al. Development and Validation of Machine Learning Algorithms Based on Electrocardiograms for Cardiovascular Diagnoses at the Population Level. npj Digit. Med. 2024, 7, 133. [Google Scholar] [CrossRef]
  21. Ose, B.; Sattar, Z.; Gupta, A.; Toquica, C.; Harvey, C.; Noheria, A. Artificial Intelligence Interpretation of the Electrocardiogram: A State-of-the-Art Review. Curr. Cardiol. Rep. 2024, 26, 561–580. [Google Scholar] [CrossRef]
  22. Somani, S.S.; Honarvar, H.; Narula, S.; Landi, I.; Lee, S.; Khachatoorian, Y.; Rehmani, A.; Kim, A.; Freitas, J.K.D.; Teng, S.; et al. Development of a Machine Learning Model Using Electrocardiogram Signals to Improve Acute Pulmonary Embolism Screening. Eur. Heart J. Digit. Health 2021, 3, 56–66. [Google Scholar] [CrossRef] [PubMed]
  23. Silva, B.V.; Marques, J.; Menezes, M.N.; Oliveira, A.L.; Pinto, F.J. Artificial Intelligence-Based Diagnosis of Acute Pulmonary Embolism: Development of a Machine Learning Model Using 12-Lead Electrocardiogram. Rev. Port. Cardiol. 2023, 42, 643–651. [Google Scholar] [CrossRef] [PubMed]
  24. Wysokinski, W.E.; Meverden, R.A.; Lopez-Jimenez, F.; Harmon, D.M.; Inojosa, B.J.M.; Suarez, A.B.; Liu, K.; Inojosa, J.R.M.; Casanegra, A.I.; McBane, R.D.; et al. Electrocardiogram Signal Analysis with a Machine Learning Model Predicts the Presence of Pulmonary Embolism with Accuracy Dependent on Embolism Burden. Mayo Clin. Proc. Digit. Health 2024, in press. [Google Scholar] [CrossRef]
  25. Vieillard-Baron, A.; Matthay, M.; Teboul, J.L.; Bein, T.; Schultz, M.; Magder, S.; Marini, J.J. Experts’ Opinion on Management of Hemodynamics in ARDS Patients: Focus on the Effects of Mechanical Ventilation. Intensive Care Med. 2016, 42, 739–749. [Google Scholar] [CrossRef] [PubMed]
  26. Ghignone, M.; Girling, L.; Prewitt, R.M. Volume Expansion versus Norepinephrine in Treatment of a Low Cardiac Output Complicating an Acute Increase in Right Ventricular Afterload in Dogs. Anesthesiology 1984, 60, 132–135. [Google Scholar] [CrossRef] [PubMed]
  27. Schneider, A.J.; Teule, G.J.J.; Groeneveld, A.B.J.; Nauta, J.; Heidendal, G.A.K.; Thijs, L.G. Biventricular Performance during Volume Loading in Patients with Early Septic Shock, with Emphasis on the Right Ventricle: A Combined Hemodynamic and Radionuclide Study. Am. Heart J. 1988, 116, 103–112. [Google Scholar] [CrossRef] [PubMed]
  28. Dessap, A.M.; Boissier, F.; Charron, C.; Bégot, E.; Repessé, X.; Legras, A.; Brun-Buisson, C.; Vignon, P.; Vieillard-Baron, A. Acute Cor Pulmonale during Protective Ventilation for Acute Respiratory Distress Syndrome: Prevalence, Predictors, and Clinical Impact. Intensive Care Med. 2016, 42, 862–870. [Google Scholar] [CrossRef]
  29. Ghio, S.; Gavazzi, A.; Campana, C. Independent and Additive Prognostic Value of Right Ventricular Systolic Function and Pulmonary Artery Pressure in Patients with Chronic Heart Failure. ACC Curr. J. Rev. 2001, 10, 56. [Google Scholar] [CrossRef]
  30. Giusca, S.; Jurcut, R.; Ticulescu, R.; Dumitru, D.; Vladaia, A.; Savu, O.; Voican, A.; Popescu, B.A.; Ginghina, C. Accuracy of Handheld Echocardiography for Bedside Diagnostic Evaluation in a Tertiary Cardiology Center: Comparison with Standard Echocardiography. Echocardiography 2011, 28, 136–141. [Google Scholar] [CrossRef]
  31. Squizzato, A.; Galli, L.; Gerdes, V.E.A. Point-of-Care Ultrasound in the Diagnosis of Pulmonary Embolism. Crit. Ultrasound J. 2015, 7, 7. [Google Scholar] [CrossRef]
  32. Cardim, N.; Dalen, H.; Voigt, J.-U.; Ionescu, A.; Price, S.; Neskovic, A.N.; Edvardsen, T.; Galderisi, M.; Sicari, R.; Donal, E.; et al. The Use of Handheld Ultrasound Devices: A Position Statement of the European Association of Cardiovascular Imaging (2018 Update). Eur. Heart J. Cardiovasc. Imaging 2019, 20, 245–252. [Google Scholar] [CrossRef] [PubMed]
  33. Daley, J.I.; Dwyer, K.H.; Grunwald, Z.; Shaw, D.L.; Stone, M.B.; Schick, A.; Vrablik, M.; Hall, M.K.; Hall, J.; Liteplo, A.S.; et al. Increased Sensitivity of Focused Cardiac Ultrasound for Pulmonary Embolism in Emergency Department Patients with Abnormal Vital Signs. Acad. Emerg. Med. 2019, 26, 1211–1220. [Google Scholar] [CrossRef] [PubMed]
Jcm 13 04792 g001
Figure 1. Patient selection flowchart.
Figure 1. Patient selection flowchart.
Jcm 13 04792 g001
Jcm 13 04792 g002
Figure 2. QCG score difference by RV dysfunction group.
Figure 2. QCG score difference by RV dysfunction group.
Jcm 13 04792 g002
Jcm 13 04792 g003
Figure 3. Performance of QCG-RVDys, Troponin I, ProBNP, and clinical experts on identifying RV dysfunction from ECGs (black line, QCG-RVDys; red line, Troponin I; blue line, ProBNP; green hollow dots, experts; black, red, and blue dots indicate binarized conditions).
Figure 3. Performance of QCG-RVDys, Troponin I, ProBNP, and clinical experts on identifying RV dysfunction from ECGs (black line, QCG-RVDys; red line, Troponin I; blue line, ProBNP; green hollow dots, experts; black, red, and blue dots indicate binarized conditions).
Jcm 13 04792 g003
Table 1. Patient characteristics.
Table 1. Patient characteristics.
Presence of RV Dysfunction
No (N = 81)Yes (N = 35)
DemographicsAge, years, median (IQR)72.0 (50.0–81.0)67.0 (45.5–77.0)0.388
Sex, male, N (%)35 (43.2%)14 (40.0%)0.907
Weight, kilograms, median (IQR)61.3 (50.2–73.8)55.8 (53.0–74.0)0.946
Height, centimeters, median (IQR)160.0 (155.0–170.0)162.0 (157.0–168.0)0.781
Risk Factors of PTEDiabetes Mellitus, N (%)18 (22.2%)5 (14.3%)0.465
Hypertension, N (%)37 (45.7%)12 (34.3%)0.350
Coronary artery occlusive disease, N (%)7 (8.6%)3 (8.6%)1.000
Cerebrovascular disease, N (%)10 (12.3%)1 (2.9%)0.209
Current smoker, N (%)5 (6.2%)3 (8.6%)0.945
Prolonged immobility (>1 week), N (%)18 (22.2%)8 (22.9%)1.000
Recent trauma or surgery (within 3 months), N (%)16 (19.8%)13 (37.1%)0.080
Active malignancy, N (%)18 (22.2%)3 (8.6%)0.136
Infectious disease (within 3 months), N (%)19 (23.5%)3 (8.6%)0.105
Hormone treatment, N (%)2 (2.5%)1 (2.9%)1.000
History of pulmonary thromboembolism, N (%)10 (12.3%)4 (11.4%)1.000
History of deep vein thrombosis, N (%)0 (0.0%)3 (8.6%)0.042
Vital SignsSystolic blood pressure, mean (SD)130.0 (26.1)129.4 (26.6)0.915
Diastolic blood pressure, mean (SD)82.2 (17.7)81.0 (15.1)0.734
Pulse rate, median (IQR) 99.0 (81.5–115.5)101.0 (90.0–117.0)0.150
Respiratory rate, median (IQR)20.0 (18.0–22.0)21.0 (20.0–24.0)0.092
Laboratory MeasurementsWhite blood cell, 109/L, median (IQR)9.2 (7.4–13.1)10.0 (7.3–13.5)0.740
Hemoglobin, g/dL, mean (SD)12.1 (2.4)12.4 (2.3)0.486
Aspartate transaminase, U/L, median (IQR)22.5 (16.0–35.0)32.0 (25.0–56.0)0.003
Alanine transaminase, U/L, median (IQR)19.0 (12.0–34.0)30.0 (18.0–60.5)0.017
Blood urea nitrogen, mg/dL, median (IQR)14.9 (11.0–22.4)14.3 (11.5–19.6)0.551
Creatinine, mg/dL, median (IQR)0.9 (0.7–1.1)0.8 (0.7–1.1)0.440
Troponin I, μg/mL, median (IQR)0.052 (0.019–0.266)0.176 (0.075–0.723)0.034
ProBNP, pg/mL, median (IQR)482.0 (140.0–2400.0)1366.0 (576.0–4733.0)0.024
D-dimer, mg/L, median (IQR)3.6 (2.1–12.3)4.6 (2.8–10.7)0.599
Lactate, mg/dL, median (IQR)1.0 (0.8–2.3)2.0 (1.1–2.3)0.097
Heart rhythm (on ECG) 0.117
Sinus Rhythm, N (%)52 (64.2%)14 (41.2%)
Sinus Tachycardia, N (%)22 (27.2%)16 (47.1%)
Atrial Fibrillation, N (%)3 (3.7%)1 (2.9%)
Multifocal Atrial Tachycardia, N (%)2 (2.5%)0 (0.0%)
Sinus Arrhythmia, N (%)1 (1.2%)0 (0.0%)
Atrial Rhythm, N (%)0 (0.0%)1 (2.9%)
Wandering Atrial Rhythm, N (%)0 (0.0%)1 (2.9%)
Undetermined Rhythm, N (%)1 (1.2%)1 (2.9%)
Right Ventricular Systolic Pressure (RVSP) <0.001
RVSP I (<35 mmHg), N (%)61 (75.3%)4 (11.4%)
RVSP II (35–49 mmHg), N (%)15 (18.5%)12 (34.3%)
RVSP III (50–64 mmHg), N (%)5 (6.2%)10 (28.6%)
RVSP IV (>64 mmHg), N (%)0 (0.0%)9 (25.7%)
Time of the testED arrival to ECG, hours, median (IQR)1.0 (0.6–1.6)0.8 (0.6–1.3)0.388
ED arrival to echocardiography, hours, median (IQR)22.1 (16.0–35.6)16.1 (4.8–27.5)0.036
ECG to echocardiography, hours, median (IQR)20.4 (14.9–34.0)13.8 (4.3–26.1)0.045
RV, right ventricular; IQR, interquartile range; PTE, pulmonary thromboembolism; SD, standard deviation; BNP, blood natriuretic peptide; ECG, electrocardiogram; ED, emergency department.
Table 2. QCG score difference by RV dysfunction and across RVSP groups.
Table 2. QCG score difference by RV dysfunction and across RVSP groups.
QCG BiomarkerGroupBiomarker Measurements, Median (IQR)p
QCG-RVDysRVD6.8 (2.5–22.5)<0.001 (for difference)
No RVD78.7 (35.7–94.9)
QCG-PHTNRVSP I (<35 mmHg)8.1 (2.1–22.9)<0.001 (for both difference and trend)
RVSP II (35–49 mmHg)21.7 (13.4–38.3)
RVSP III (50–64 mmHg)41.6 (31.2–65.1)
RVSP IV (>64 mmHg)49.2 (19.9–85.4)
QCG, quantitative electrocardiography; IQR, interquartile range; RVDys, right ventricular dysfunction; RVD, right ventricular dysfunction; PHTN, pulmonary hypertension; RVSP, right ventricular systolic pressure.
Table 3. Performance of QCG and other biomarkers for prediction of RV dysfunction.
Table 3. Performance of QCG and other biomarkers for prediction of RV dysfunction.
BiomarkerAUCp for DifferenceSensitivity
(95% CI)		Specificity
(95% CI)		PPV
(95% CI)		NPV
(95% CI)		Threshold
QCG-RVDys (continuous scale)0.895 (0.829–0.960)-91.2
(82.4–100)		77.8
(69.1–86.4)		63.3
(54.4–73.9)		95.5
(90.8–100)		24.65
Troponin I0.692 (0.536–0.847)0.04681.2
(62.5–100)		54.8
(38.7–71)		48.1
(37.5–61.1)		85
(70.6–100)		0.0685 μg/mL
ProBNP0.655 (0.532–0.778)0.00180.8
(65.4–96.2)		53.4
(39.7–65.5)		43.8
(35.6–52.5)		86.4
(75.8–96.3)		534.5 pg/mL
AUC, area under the curve; PPV, positive predictive value; CI, confidence interval; NPV, negative predictive value; QCG, quantitative electrocardiography; RVDys, right ventricular dysfunction; BNP, blood natriuretic peptide.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Choi, Y.J.; Park, M.J.; Cho, Y.; Kim, J.; Lee, E.; Son, D.; Kim, S.-Y.; Soh, M.S. Screening for RV Dysfunction Using Smartphone ECG Analysis App: Validation Study with Acute Pulmonary Embolism Patients. J. Clin. Med. 2024, 13, 4792. https://doi.org/10.3390/jcm13164792

AMA Style

Choi YJ, Park MJ, Cho Y, Kim J, Lee E, Son D, Kim S-Y, Soh MS. Screening for RV Dysfunction Using Smartphone ECG Analysis App: Validation Study with Acute Pulmonary Embolism Patients. Journal of Clinical Medicine. 2024; 13(16):4792. https://doi.org/10.3390/jcm13164792

Chicago/Turabian Style

Choi, Yoo Jin, Min Ji Park, Youngjin Cho, Joonghee Kim, Eunkyoung Lee, Dahyeon Son, Seo-Yoon Kim, and Moon Seung Soh. 2024. "Screening for RV Dysfunction Using Smartphone ECG Analysis App: Validation Study with Acute Pulmonary Embolism Patients" Journal of Clinical Medicine 13, no. 16: 4792. https://doi.org/10.3390/jcm13164792

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop