Cardiac resynchronization therapy (CRT) is a device intervention for heart failure patients. While current criteria for CRT eligibility depend on professional society guidelines, it is generally indicated for heart failure patients with a left bundle branch block (LBBB). However, LBBB is not uniformly defined and the more recently proposed “strict” LBBB criteria have been proposed as a better predictor of benefit from CRT. Automatic detection of “strict” LBBB criteria can improve outcomes for heart failure patient by reducing high false positive rates in LBBB detection.
This study proposes an algorithm to automatically detect strict LBBB, developed and tested using ECGs made available via the International Society of Computerized Electrocardiology LBBB initiative. The dataset consists of 12-lead Holter ECGs recorded before therapy from the MADIT-CRT clinical trial.
The algorithm consists of multi-lead QRS complex detection using length transform and a support vector machine (SVM) classifier to identify QS- and rS- configurations using mean power, maximum/minimum amplitude and signal entropy as features. Notching is identified by the inflection points of the signal which could be extracted from the first derivative of the signal. Similarly, slurring is detected by the inflection points of the first derivative which is related to the second derivative of the signal. The algorithm detects both the presence of notches and slurs and time of onset and offset which is vital for “strict” LBBB criteria.
The algorithm achieved a sensitivity of 79\%, specificity of 77\% and positive predictive value (PPV) of 63\% on the training set. It achieved sensitivity, specificity and PPV of 64\%, 73\% and 53\% on the hidden test set. One of the drawbacks of the algorithm is the high sensitivity to minor slurring. Such slurs met the inflection point detection criterion used to detect slurs which results in low PPV and specificity for LBBB detection.