Identifying Ventricular Dysfunction Indicators in Electrocardiograms via Artificial Intelligence-Driven Analysis
Abstract
1. Introduction
2. Methods
2.1. Dataset Preparation
2.2. Data Preprocessing
- (1)
- Three-second ECG segmentation: Each record was segmented into multiple 3 s intervals as part of an oversampling strategy based on predefined time steps aligned with the LVEF (Figure 1). In the training datasets, the oversampling factors were x4 for preserved EF (pEF), x7 for mrEF, and x6 for rEF. For the development and test datasets, the oversampling factors were x4 for pEF, x16 for mrEF, and x11 for rEF. The segmented 3 s ECG data were then exported as CSV files, each containing 12 rows and 1500 columns.
- (2)
- Single-beat ECG extraction: Using Ngaia™ from Nagaoka Industries (Nishinomiya, Japan), ECG data were segmented based on the detected QRS complex. Data spanning 250 ms before and 500 ms after the QRS complex were extracted. The data extended to the end of the T-wave and were replaced with the value at the end of the T-wave. These segmented single-beat ECG data were then exported as CSV files, each containing 12 rows and 375 columns.
- (3)
- Partial ECG: From a single-beat ECG, five types of partial ECGs were created, demarcated by the data points 100 ms before the QRS, at the ST junction, and at the end of the T-wave, including ① P, ② PQRS, ③ QRS, ④ QRST, ⑤ PQRST (Figure 2). Based on the single-beat ECG data, the partial ECGs were segmented as follows: the boundary between the P-wave and QRS complex was defined as 100 ms before the detected QRS complex. The end of the QRS complex and the T-wave was also detected using the Ngaia™ programme.
- (4)
- Two-beat ECG composition: Using the same Ngaia™ software, two-beat ECGs were constructed by combining subsequent QRS complex channel-wise. This resulted in a dataset with 24 channels composed of 24 rows and 375 columns in CSV files. The last QRS complex of a 10 s ECG, lacking a subsequent QRS complex, was combined with the first QRS complex of the same interval.
- (5)
- Reduced lead ECG: ECGs recorded in 12 leads were divided into individual lead data. Selected rows from the single-beat ECG data were utilised, resulting in datasets comprising between 1 and 12 rows and 375 columns. The total number of lead combinations was 40,955 (12C1 + 12C2 + … + 12C12).
2.3. Architecture of AI
2.4. Model Development and Validation
3. Results
3.1. Comparison of 3 s ECG, Single-Beat ECG, and Two-Beat ECG Models for Detecting Ventricular Dysfunction Subsection
3.2. Effect of Lead Reduction and Lead Configuration
3.3. Comparison of Partial ECG Models
4. Discussion
5. Conclusions
Supplementary Materials
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Multimodality Writing Group for Chronic Coronary Disease. ACC/AHA/ASE/ASNC/ASPC/HFSA/HRS/SCAI/SCCT/SCMR/STS 2023 multimodality appropriate use criteria for the detection and risk assessment of chronic coronary disease. J. Am. Coll. Cardiol. 2023, 81, 2445–2467. [Google Scholar] [CrossRef] [PubMed]
- McDonagh, T.A.; Metra, M.; Adamo, M.; Gardner, R.S.; Baumbach, A.; Böhm, M.; Burri, H.; Butler, J.; Čelutkienė, J.; Chioncel, O.; et al. 2023 Focused Update of the 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur. Heart J. 2023, 44, 3627–3639. [Google Scholar] [CrossRef] [PubMed]
- 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]
- Harmon, D.M.; Carter, R.E.; Cohen-Shelly, M.; Svatikova, A.; Adedinsewo, D.A.; Noseworthy, P.A.; Kapa, S.; Lopez-Jimenez, F.; Friedman, P.A.; Attia, Z.I. Real-world performance, long-term efficacy, and absence of bias in the artificial intelligence enhanced electrocardiogram to detect left ventricular systolic dysfunction. Eur. Heart J. Digit. Health 2022, 3, 238–244. [Google Scholar] [CrossRef] [PubMed]
- Ohte, N.; Ishizu, T.; Izumi, C.; Itoh, H.; Iwanaga, S.; Okura, H.; Otsuji, Y.; Sakata, Y.; Shibata, T.; Shinke, T.; et al. JCS 2021 guideline on the clinical application of echocardiography. Circ. J. 2022, 86, 2045–2119. [Google Scholar] [CrossRef]
- Popescu, B.A.; Stefanidis, A.; Nihoyannopoulos, P.; Fox, K.F.; Ray, S.; Cardim, N.; Rigo, F.; Badano, L.P.; Fraser, A.G.; Pinto, F.; et al. Updated standards and processes for accreditation of echocardiographic laboratories from The European Association of Cardiovascular Imaging. Eur. Heart J. Cardiovasc. Imaging 2014, 15, 717–727. [Google Scholar] [CrossRef]
- He, K.; Zhang, X.; Ren, S.; Sun, J. Delving deep into rectifiers: Surpassing human-level performance on imagenet classification. In Proceedings of the 2015 IEEE International Conference on Computer Vision (ICCV), Santiago, Chile, 7–13 December 2015. [Google Scholar]
- Tan, T.; Yin, S.K.; Liu, K.; Wan, M. On the convergence speed of AMSGRAD and Beyond. In Proceedings of the 31st IEEE International Conference on Tools with Artificial Intelligence, ICTAI 2019, Portland, OR, USA, 4–6 November 2019. [Google Scholar]
- Zhang, H.; Cisse, M.; Dauphin, Y.N.; Lopez-Paz, D. mixup: Beyond Empirical Risk Minimization. arXiv 2017, arXiv:1710.09412. [Google Scholar]
- Attia, Z.I.; Kapa, S.; Lopez-Jimenez, F.; McKie, P.M.; Ladewig, D.J.; Satam, G.; Pellikka, P.A.; Enriquez-Sarano, M.; Noseworthy, P.A.; Munger, T.M.; et al. Screening for cardiac contractile dysfunction using an artificial intelligence-enabled electrocardiogram. Nat. Med. 2019, 25, 70–74. [Google Scholar] [CrossRef]
- König, S.; Hohenstein, S.; Nitsche, A.; Pellissier, V.; Leiner, J.; Stellmacher, L.; Hindricks, G.; Bollmann, A. Artificial intelligence-based identification of left ventricular systolic dysfunction from 12-lead electrocardiograms: External validation and advanced application of an existing model. Eur. Heart J. Digit. Health 2023, 5, 144–151. [Google Scholar] [CrossRef]
- Kim, Y.; Lee, M.; Yoon, J.; Kim, Y.; Min, H.; Cho, H.; Park, J.; Shin, T. Predicting future incidences of cardiac arrhythmias using discrete heartbeats from normal sinus rhythm ECG signals via deep learning methods. Diagnostics 2023, 13, 2849. [Google Scholar] [CrossRef]
- Deevi, S.A.; Kaniraja, C.P.; Mani, V.D.; Mishra, D.; Ummar, S.; Satheesh, C. HeartNetEC: A deep representation learning approach for ECG beat classification. Biomed. Eng. Lett. 2021, 11, 69–84. [Google Scholar] [CrossRef] [PubMed]
- Obeidat, Y.; Alqudah, A.M. A hybrid lightweight 1D CNN-LSTM architecture for automated ECG beat-wise classification. Trait. Signal 2021, 38, 1281–1291. [Google Scholar] [CrossRef]
- Riera, A.R.; Ferreira, C.; Ferreira Filho, C.; Dubner, S.; Barbosa Barros, R.; Femenía, F.; Baranchuk, A. Clinical value of lead aVR. Ann. Noninvasive Electrocardiol. 2011, 16, 295–302. [Google Scholar] [CrossRef]
- Kireyev, D.; Arkhipov, M.V.; Zador, S.T.; Paris, J.A.; Boden, W.E. Clinical utility of aVR-The neglected electrocardiographic lead. Ann. Noninvasive Electrocardiol. 2010, 15, 175–180. [Google Scholar] [CrossRef]
- Pahlm, U.S.; Pahlm, O.; Wagner, G.S. The standard 11-lead ECG. Neglect of lead aVR in the classical limb lead display. J. Electrocardiol. 1996, 29 (Suppl. S1), 270. [Google Scholar] [CrossRef]
- Selvester, R.H.; Wagner, G.S.; Ideker, R.E.; Gates, K.; Starr, S.; Ahmed, J.; Crump, R. ECG myocardial infarct size: A gender-, age-, race-insensitive 12-segment multiple regression model. I: Retrospective learning set of 100 pathoanatomic infarcts and 229 normal control subjects. J. Electrocardiol. 1994, 27 (Suppl. S1), 31. [Google Scholar] [CrossRef]
- Russo, G.; Ravenna, S.E.; De Vita, A.; Aurigemma, C.; Lamendola, P.; Lanza, G.A.; Crea, F. Exercise test predictors of severe coronary artery disease: Role of ST-segment elevation in lead aVR. Clin. Cardiol. 2017, 40, 102. [Google Scholar] [CrossRef] [PubMed]
- Glöckner, D.; Zeus, T.; Kelm, M.; Makimoto, H. Recurrent nocturnal ST-T deviation and nonsustained ventricular tachycardias recorded with a smartwatch: A case report. J. Cardiol. Cases 2022, 26, 169–172. [Google Scholar] [CrossRef]
- Attia, Z.I.; Harmon, D.M.; Dugan, J.; Manka, L.; Lopez-Jimenez, F.; Lerman, A.; Siontis, K.C.; Noseworthy, P.A.; Yao, X.; Klavetter, E.W.; et al. Prospective evaluation of smartwatch-enabled detection of left ventricular dysfunction. Nat. Med. 2022, 28, 2497–2503. [Google Scholar] [CrossRef]
- Haslam, D.W.; James, W.P. Obesity. Lancet 2005, 366, 1197–1209. [Google Scholar] [CrossRef]
- Hunma, S.; Ramuth, H.; Miles-Chan, J.L.; Schutz, Y.; Montani, J.P.; Joonas, N.; Dulloo, A.G. Body composition-derived BMI cut-offs for overweight and obesity in Indians and Creoles of Mauritius: Comparison with Caucasians. Int. J. Obes. 2016, 40, 1906–1914. [Google Scholar] [CrossRef] [PubMed]
- Makimoto, H.; Höckmann, M.; Lin, T.; Glöckner, D.; Gerguri, S.; Clasen, L.; Schmidt, J.; Assadi-Schmidt, A.; Bejinariu, A.; Müller, P.; et al. Performance of a convolutional neural network derived from an ECG database in recognizing myocardial infarction. Sci. Rep. 2020, 10, 8445. [Google Scholar] [CrossRef] [PubMed]
- Makimoto, H.; Shiraga, T.; Kohlmann, B.; Magnisali, C.E.; Gerguri, S.; Motoyama, N.; Clasen, L.; Bejinariu, A.; Klein, K.; Makimoto, A.; et al. Efficient screening for severe aortic valve stenosis using understandable artificial intelligence: A prospective diagnostic accuracy study. Eur. Heart J. Digit. Health 2022, 3, 141–152. [Google Scholar] [CrossRef] [PubMed]
- Shiraga, T.; Makimoto, H.; Kohlmann, B.; Magnisali, C.E.; Imai, Y.; Itani, Y.; Makimoto, A.; Schölzel, F.; Bejinariu, A.; Kelm, M.; et al. Improving valvular pathologies and ventricular dysfunction diagnostic efficiency using combined auscultation and electrocardiography data: A multimodal AI approach. Sensors 2023, 23, 9834. [Google Scholar] [CrossRef]
Institution | Number of Unique Cases | Number of ECGs | LVEF | For Train and Dev | For Test |
---|---|---|---|---|---|
JMU | 10,626 | 17,021 | pEF | 10,398 | 2677 |
mrEF | 1452 | 387 | |||
rEF | 1709 | 398 | |||
UHD | 401 | 401 | pEF | 190 | |
mrEF | 159 | ||||
rEF | 52 |
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. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Makimoto, H.; Okatani, T.; Suganuma, M.; Kabutoya, T.; Kohro, T.; Agata, Y.; Ogata, Y.; Harada, K.; Llubani, R.; Bejinariu, A.; et al. Identifying Ventricular Dysfunction Indicators in Electrocardiograms via Artificial Intelligence-Driven Analysis. Bioengineering 2024, 11, 1069. https://doi.org/10.3390/bioengineering11111069
Makimoto H, Okatani T, Suganuma M, Kabutoya T, Kohro T, Agata Y, Ogata Y, Harada K, Llubani R, Bejinariu A, et al. Identifying Ventricular Dysfunction Indicators in Electrocardiograms via Artificial Intelligence-Driven Analysis. Bioengineering. 2024; 11(11):1069. https://doi.org/10.3390/bioengineering11111069
Chicago/Turabian StyleMakimoto, Hisaki, Takayuki Okatani, Masanori Suganuma, Tomoyuki Kabutoya, Takahide Kohro, Yukiko Agata, Yukiyo Ogata, Kenji Harada, Redi Llubani, Alexandru Bejinariu, and et al. 2024. "Identifying Ventricular Dysfunction Indicators in Electrocardiograms via Artificial Intelligence-Driven Analysis" Bioengineering 11, no. 11: 1069. https://doi.org/10.3390/bioengineering11111069
APA StyleMakimoto, H., Okatani, T., Suganuma, M., Kabutoya, T., Kohro, T., Agata, Y., Ogata, Y., Harada, K., Llubani, R., Bejinariu, A., Rana, O. R., Makimoto, A., Gharib, E., Meissner, A., Kelm, M., & Kario, K. (2024). Identifying Ventricular Dysfunction Indicators in Electrocardiograms via Artificial Intelligence-Driven Analysis. Bioengineering, 11(11), 1069. https://doi.org/10.3390/bioengineering11111069