Zeineb Bouzid1, Ziad Faramand2, Richard E Gregg3, Stephanie Helman4, Christian Martin-Gill5, Samir Saba6, Clifton Callaway5, Ervin Sejdić7, Salah Al-Zaiti8. 1. Department of Electrical & Computer Engineering, PA, USA. 2. Department of Acute & Tertiary Care Nursing, PA, USA; University of Pittsburgh Medical Center (UPMC), Pittsburgh, PA, USA. 3. Advanced Algorithm Research Center, Philips Healthcare, Andover, MA, USA. 4. Department of Acute & Tertiary Care Nursing, PA, USA. 5. Department of Emergency Medicine, PA, USA; University of Pittsburgh Medical Center (UPMC), Pittsburgh, PA, USA. 6. Division of Cardiology at University of Pittsburgh, PA, USA; University of Pittsburgh Medical Center (UPMC), Pittsburgh, PA, USA. 7. Department of Electrical & Computer Engineering, PA, USA; Department of Bioengineering at Swanson School of Engineering, PA, USA; Department of Biomedical Informatics at School of Medicine, PA, USA; Intelligent Systems Program at School of Computing and Information, PA, USA. 8. Department of Acute & Tertiary Care Nursing, PA, USA; Department of Emergency Medicine, PA, USA; Division of Cardiology at University of Pittsburgh, PA, USA. Electronic address: ssa33@pitt.edu.
Abstract
BACKGROUND: Novel temporal-spatial features of the 12‑lead ECG can conceptually optimize culprit lesions' detection beyond that of classical ST amplitude measurements. We sought to develop a data-driven approach for ECG feature selection to build a clinically relevant algorithm for real-time detection of culprit lesion. METHODS: This was a prospective observational cohort study of chest pain patients transported by emergency medical services to three tertiary care hospitals in the US. We obtained raw 10-s, 12‑lead ECGs (500 s/s, HeartStart MRx, Philips Healthcare) during prehospital transport and followed patients 30 days after the encounter to adjudicate clinical outcomes. A total of 557 global and lead-specific features of P-QRS-T waveform were harvested from the representative average beats. We used Recursive Feature Elimination and LASSO to identify 35/557, 29/557, and 51/557 most recurrent and important features for LAD, LCX, and RCA culprits, respectively. Using the union of these features, we built a random forest classifier with 10-fold cross-validation to predict the presence or absence of culprit lesions. We compared this model to the performance of a rule-based commercial proprietary software (Philips DXL ECG Algorithm). RESULTS: Our sample included 2400 patients (age 59 ± 16, 47% female, 41% Black, 10.7% culprit lesions). The area under the ROC curves of our random forest classifier was 0.85 ± 0.03 with sensitivity, specificity, and negative predictive value of 71.1%, 84.7%, and 96.1%. This outperformed the accuracy of the automated interpretation software of 37.2%, 95.6%, and 92.7%, respectively, and corresponded to a net reclassification improvement index of 23.6%. Metrics of ST80; Tpeak-Tend; spatial angle between QRS and T vectors; PCA ratio of STT waveform; T axis; and QRS waveform characteristics played a significant role in this incremental gain in performance. CONCLUSIONS: Novel computational features of the 12‑lead ECG can be used to build clinically relevant machine learning-based classifiers to detect culprit lesions, which has important clinical implications.
BACKGROUND: Novel temporal-spatial features of the 12‑lead ECG can conceptually optimize culprit lesions' detection beyond that of classical ST amplitude measurements. We sought to develop a data-driven approach for ECG feature selection to build a clinically relevant algorithm for real-time detection of culprit lesion. METHODS: This was a prospective observational cohort study of chest pain patients transported by emergency medical services to three tertiary care hospitals in the US. We obtained raw 10-s, 12‑lead ECGs (500 s/s, HeartStart MRx, Philips Healthcare) during prehospital transport and followed patients 30 days after the encounter to adjudicate clinical outcomes. A total of 557 global and lead-specific features of P-QRS-T waveform were harvested from the representative average beats. We used Recursive Feature Elimination and LASSO to identify 35/557, 29/557, and 51/557 most recurrent and important features for LAD, LCX, and RCA culprits, respectively. Using the union of these features, we built a random forest classifier with 10-fold cross-validation to predict the presence or absence of culprit lesions. We compared this model to the performance of a rule-based commercial proprietary software (Philips DXL ECG Algorithm). RESULTS: Our sample included 2400 patients (age 59 ± 16, 47% female, 41% Black, 10.7% culprit lesions). The area under the ROC curves of our random forest classifier was 0.85 ± 0.03 with sensitivity, specificity, and negative predictive value of 71.1%, 84.7%, and 96.1%. This outperformed the accuracy of the automated interpretation software of 37.2%, 95.6%, and 92.7%, respectively, and corresponded to a net reclassification improvement index of 23.6%. Metrics of ST80; Tpeak-Tend; spatial angle between QRS and T vectors; PCA ratio of STT waveform; T axis; and QRS waveform characteristics played a significant role in this incremental gain in performance. CONCLUSIONS: Novel computational features of the 12‑lead ECG can be used to build clinically relevant machine learning-based classifiers to detect culprit lesions, which has important clinical implications.
Authors: C P Cannon; A Battler; R G Brindis; J L Cox; S G Ellis; N R Every; J T Flaherty; R A Harrington; H M Krumholz; M L Simoons; F J Van De Werf; W S Weintraub; K R Mitchell; S L Morrisson; R G Brindis; H V Anderson; D S Cannom; W R Chitwood; J E Cigarroa; R L Collins-Nakai; S G Ellis; R J Gibbons; F L Grover; P A Heidenreich; B K Khandheria; S B Knoebel; H L Krumholz; D J Malenka; D B Mark; C R Mckay; E R Passamani; M J Radford; R N Riner; J B Schwartz; R E Shaw; R J Shemin; D B Van Fossen; E D Verrier; M W Watkins; D R Phoubandith; T Furnelli Journal: J Am Coll Cardiol Date: 2001-12 Impact factor: 24.094
Authors: Patrick T O'Gara; Frederick G Kushner; Deborah D Ascheim; Donald E Casey; Mina K Chung; James A de Lemos; Steven M Ettinger; James C Fang; Francis M Fesmire; Barry A Franklin; Christopher B Granger; Harlan M Krumholz; Jane A Linderbaum; David A Morrow; L Kristin Newby; Joseph P Ornato; Narith Ou; Martha J Radford; Jacqueline E Tamis-Holland; Carl L Tommaso; Cynthia M Tracy; Y Joseph Woo; David X Zhao Journal: J Am Coll Cardiol Date: 2012-12-17 Impact factor: 24.094
Authors: Ezra A Amsterdam; Nanette K Wenger; Ralph G Brindis; Donald E Casey; Theodore G Ganiats; David R Holmes; Allan S Jaffe; Hani Jneid; Rosemary F Kelly; Michael C Kontos; Glenn N Levine; Philip R Liebson; Debabrata Mukherjee; Eric D Peterson; Marc S Sabatine; Richard W Smalling; Susan J Zieman Journal: Circulation Date: 2014-09-23 Impact factor: 29.690
Authors: Ivo Strebel; Raphael Twerenbold; Jasper Boeddinghaus; Roger Abächerli; Maria Rubini Giménez; Karin Wildi; Karin Grimm; Christian Puelacher; Patrick Badertscher; Zaid Sabti; Dominik Breitenbücher; Janina Jann; Farah Selman; Jeanne du Fay de Lavallaz; Nicolas Schaerli; Thomas Nestelberger; Claudia Stelzig; Michael Freese; Lukas Schumacher; Stefan Osswald; Christian Mueller; Tobias Reichlin Journal: Ann Noninvasive Electrocardiol Date: 2018-02-24 Impact factor: 1.468
Authors: Roger Abächerli; Raphael Twerenbold; Jasper Boeddinghaus; Thomas Nestelberger; Patrick Mächler; Roberto Sassi; Massimo W Rivolta; Ebadollah Kheirati Roonizi; Luca T Mainardi; Nikola Kozhuharov; Maria Rubini Giménez; Karin Wildi; Karin Grimm; Zaid Sabti; Petra Hillinger; Christian Puelacher; Ivo Strebel; Janosch Cupa; Patrick Badertscher; Isabelle Roux; Ramun Schmid; Remo Leber; Stefan Osswald; Christian Mueller; Tobias Reichlin Journal: Int J Cardiol Date: 2017-02-07 Impact factor: 4.164
Authors: Ivo Strebel; Raphael Twerenbold; Desiree Wussler; Jasper Boeddinghaus; Thomas Nestelberger; Jeanne du Fay de Lavallaz; Roger Abächerli; Patrick Maechler; Diego Mannhart; Nikola Kozhuharov; Maria Rubini Giménez; Karin Wildi; Lorraine Sazgary; Zaid Sabti; Christian Puelacher; Patrick Badertscher; Dagmar I Keller; Òscar Miró; Carolina Fuenzalida; Sofia Calderón; F Javier Martin-Sanchez; Sergio Lopez Iglesias; Stefan Osswald; Christian Mueller; Tobias Reichlin Journal: Int J Cardiol Date: 2018-09-19 Impact factor: 4.164
Authors: Salah S Al-Zaiti; Christian Martin-Gill; Ervin Sejdić; Mohammad Alrawashdeh; Clifton Callaway Journal: J Electrocardiol Date: 2015-08-06 Impact factor: 1.438
Authors: Salah S Al-Zaiti; Clifton W Callaway; Teri M Kozik; Mary G Carey; Michele M Pelter Journal: J Am Heart Assoc Date: 2015-07-24 Impact factor: 5.501