Feature Engineering and Machine Learning for the Auto-Detection of Misplaced V1 and V2 Chest Electrodes when Recording 12-lead ECGs

Khaled Rjoob, RR Bond, D Finlay, V. E. McGilligan, Stephen James Leslie, Aleeha Iftikhar, D Guldenring, Charles Knoery, Aaron Peace

Research output: Contribution to journalMeeting Abstractpeer-review

28 Downloads (Pure)


Background: Electrocardiogram (ECG) lead misplacement can adversely affect ECG diagnosis and subsequent clinical decisions. V1 and V2 are commonly placed too high and wide of their correct position. The aim of the current study was to use machine learning to detect lead V1 and V2 misplacement to enhance ECG data quality to improve clinical decision making in cardiac care.

Methods: ECGs for 453 patients, (normal n=151, Left Ventricular Hypertrophy (LVH) n=151, Myocardial Infarction n=151) were evaluated from Body surface potential maps. These were used to extract correct and incorrectly placed leads. The prevalence or class balance for correct and incorrect leads were 50%. Sixteen features were extracted in three different domains: time-based, statistical and time-frequency features using a wavelet transform. A hybrid feature selection approach that combines the filter method and wrapper method to select an optimal set of features that provides a good classification accuracy was applied. To ensure accuracy, five classifiers were used including: fine tree, coarse tree, Linear Support Vector Machine (LSVM), Quadratic Support Vector Machine (QSVM) and logistic regression. The aforementioned feature selection approach and classifiers were applied for V1 and V2 misplacement in three different positions: first, second and third Intercostal Spaces (ICS).

Results: The accuracy for V1 misplacement detection was 93.9%, 89.3%, 72.8%, sensitivity was 96.6%, 89.9%, 71.1% and specificity was 91.2%, 88.5%, 74.5% in the first, second and third ICS respectively. In V2, the accuracy for lead misplacement detection was 93.6%, 86.6% and 68.1%, sensitivity was 95.3%, 86.6%, 67.1% and specificity was 91.9%, 86.6%, 69.1% in the first, second and third ICS respectively.

Conclusions: Based on accuracy, sensitivity and specificity QSVM was shown to be the best classifier in V1 in the first and the second ICS combined with Joint Mutual Information (JMI) feature selection algorithm and entropy-based features sets respectively. Logistic regression was the best in the third ICS combined with Mutual Information Feature Selection (MIFS) algorithm. In V2, QSVM was the best classifier in the first and the third ICS combined with JMI and MIFS respectively, while fine tree was the best in the second ICS combined with MIFS. There is a noticeable decline in accuracy when detecting misplacement in the third ICS, so this indicates that when the electrode is gradually repositioned to the correct position, the feature values could be very close to the feature values in the correct position.
Original languageEnglish
Pages (from-to)S106
JournalJournal of Electrocardiology
Publication statusPublished (in print/issue) - 6 Dec 2019
EventInternational Society for Computerised Electrocardiology - Atlantic Beach, Jacksonville, United States
Duration: 10 Apr 201914 Dec 2019


  • Feature Engineering
  • Machine Learning
  • ECG
  • Lead misplacement


Dive into the research topics of 'Feature Engineering and Machine Learning for the Auto-Detection of Misplaced V1 and V2 Chest Electrodes when Recording 12-lead ECGs'. Together they form a unique fingerprint.

Cite this