Automatic left bundle branch block diagnose using a 2-D convolutional network

Abstract

Left bundle branch block (LBBB) patients are the populationthat benefits most from cardiac resynchronization therapy (CRT),a therapy applied in heart failure. However, CRT presents about 40%non-responders rates. A plausible explanation to this fact, is a precariousLBBB diagnosis. QRS duration is currently one of three pillars inLBBB diagnosis. However, ECG morphology is severely altered in thepresence of LBBB, affecting seriously the process of ECG delineation.Thus, QRS duration becomes a highly unreliable measure in LBBB diagnosis.Herein, we propose a LBBB classification framework complettelyindependent of temporal measures. In this line, a 2-D convolutional network(CNN) was utilized to separate strict LBBB patients from (notstrict/not) LBBB patients, obtained from a subset of the Multi-centerAutonomic Defibrillator Implantation (MADIT) trial. In order to fit the2-D architecture, we fed the CNN with 10 s- spectrograms, constructingand validating 6 separated unilead models, one per precordial lead.From all analyzed models, the one using lead V1 turned out to be themost informative. The latter, produced an 89% accuracy and 90% positivepredictive value. These results encourage the use of such statisticalmodels to provide a more reliable and automated LBBB diagnosis.