Automatic Diagnosis of Strict Left Bundle Branch Block from Standard 12-lead Electrocardiogram

Strict Left Bundle Branch Block (LBBB) criteria were recently proposed to identify patients with complete LBBB to benefit most from Cardiac Resynchronization Therapy (CRT). The objective of our study was to automate this strict LBBB criteria in order to facilitate broader application of the criteria which require the measurements of subtle QRS patterns from standard 12-lead ECGs. We developed a series of algorithms to automatically detect and measure the QRS parameters required for strict LBBB criteria. A total of 612 signal-averaged 12-lead ECGs from 612 LBBB patients were used to train and validate the algorithms. Four clinicians independently performed adjudication on equally assigned ECGs to assess the performance of automatic results comparing to manually adjudicated results, as well as the inter-observer and intra-observer variabilities. Overall 95% and 86% of sensitivity and specificity are reached for detecting complete LBBB. Our study shows good performance in reference to manual results.