Session S24.2

Enhanced Detection of Electrode Placement/Connection Errors

C Cooper, E Clark, PW Macfarlane*

University of Glasgow
Glasgow, UK

Introduction: It is quite possible that millions of ECGs recorded annually have faulty electrode connection errors. Some are more obvious than others, but automated detection can help to minimise such errors by advising the technician at the time of recording of any problem detected.
Aims: i) To determine the accuracy of the Glasgow Program in detecting lead connection or electrode placement errors; ii) To improve the performance of the program with respect to (i).
Methods: 50 volunteers were recruited from hospital patients and staff, including students. For each volunteer, a 12 lead ECG was recorded correctly, together with additional 6 recordings where, in three cases, there was a deliberate error in limb lead connections and in three other cases, there was a deliberate electrode placement error in the precordial leads. Half of the ECGs were used as a training set to assess the ability of an existing version of the program to detect such errors and then subsequently for enhancement of the relevant diagnostic criteria, while the remaining half were then used as a test set for assessing newly developed criteria.
Results: The sensitivity of detecting arm lead reversal improved from 64% to 88%, with 100% specificity. The sensitivity of detecting a right leg mis-connection improved from 84% to 96%, with almost 100% specificity. New criteria for V1 and V2 interchange had 92% sensitivity and almost 100% specificity, while criteria for V2/V3 interchange improved from 10% to 24% sensitivity at almost 100% specificity. Detection of V1/V2 placement one inter-space too high proved impossible in this particular cohort as waveforms did not change to any significant extent.
Conclusion: Simple enhancements to classical rule based criteria for the detection of electrode connection or misplacement errors resulted in a vastly improved sensitivity in most cases, except for V2/V3 interchange error, which proved extremely difficult to detect accurately while maintaining a high specificity.

(Abstract Control Number: 57)