Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Predicting Acute Myocardial Ischemia using Machine Learning applied to Standard 10-second 12-lead ECG

Besomi, Lucas (2019) Predicting Acute Myocardial Ischemia using Machine Learning applied to Standard 10-second 12-lead ECG. Master's Thesis, University of Pittsburgh. (Unpublished)

This is the latest version of this item.

[img] PDF
Restricted to University of Pittsburgh users only until 10 September 2024.

Download (2MB) | Request a Copy


Prompt identification of acute coronary syndrome (ACS) is a pivotal challenge to reduce the cost burden related to the prolonged hospitalization of patients with minor cardiovascular issues. Non-ST-segment elevation myocardial infarction and unstable angina are complex to identify and delay the diagnosis of ACS, requiring invasive and costly tests. A screening tool exploiting the different temporal-spatial features of the standard 10-second 12-lead electrocardiogram (ECG) and ruling out patients without risk would then be useful in the admission process of such patients. In this study we compared the performance of the following machine learning algorithms in detecting ACS: Gradient Boosting Machine (GBM), Logistic Regression (LogReg) and Artificial Neural Network (ANN). The first two cohorts of the ongoing EMPIRE (ECG Methods for the Prompt Identification of Coronary Events) study were used as a data set, with 750 patients for training and 500 patients for testing. Three different versions of this data set were used: the initial version with all the features, a reduced version with fewer features, and a final version with re-labeled features. The models were evaluated a first time using only ECG-based features, and then with additional clinical features in order to observe any improvement. The best results were obtained with LogReg on the reduced and labeled version of the data set, with a negative predictive value of 93.4%, a specificity of 76.6% and a sensitivity of 75.6% on the test set. No significant changes were noted after adding clinical features to our classifiers. To conclude, reduction and labeling of the data set proved to be beneficial to our model and this latter could be used as a supportive tool when performing the standard 10-second 12-lead ECG at the very early stage of the continuum of care.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Besomi, Lucasleb138@pitt.eduLEB1380000-0001-9995-133X
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairSejdic, Ervinesejdic@pitt.eduesejdic0000-0003-4987-8298
Committee MemberAkcakaya, Muratakcakaya@pitt.eduakcakaya0000-0001-5094-1931
Committee MemberEl-Jaroudi, Amroamro@pitt.eduamro
Committee MemberAl-Zaiti, SalahSSA33@pitt.eduSSA330000-0002-6862-0658
Date: 10 September 2019
Date Type: Publication
Defense Date: 12 July 2019
Approval Date: 10 September 2019
Submission Date: 12 July 2019
Access Restriction: 5 year -- Restrict access to University of Pittsburgh for a period of 5 years.
Number of Pages: 53
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Electrical and Computer Engineering
Degree: MS - Master of Science
Thesis Type: Master's Thesis
Refereed: Yes
Uncontrolled Keywords: cute myocardial ischemia, prompt identification of ACS, EMPIRE, ECG, machine learning
Date Deposited: 10 Sep 2019 14:00
Last Modified: 10 Sep 2019 14:00

Available Versions of this Item

  • Predicting Acute Myocardial Ischemia using Machine Learning applied to Standard 10-second 12-lead ECG. (deposited 10 Sep 2019 14:00) [Currently Displayed]


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item