Inappropriate surface ECG signal filtering significantly reduces physicians' ability to recognize left bundle branch block

QRS notching and slurring are key morphological markers of Strauss-defined left bundle branch block (LBBB). The study aimed to evaluate the effect of ECG filtering on QRS morphology and the diagnosis of Strauss-defined LBBB.

This retrospective study included 53 CRT candidates who underwent 12-lead ECG recording before device implantation. ECGs were processed under four filtering conditions: unfiltered, strong filtering (0.05–25 Hz), moderate filtering (0.05–60 Hz), and mild filtering (0.05–150 Hz). QRS notching, slurring, and plateau were visually assessed in leads I, aVL, V1, V2, V5, and V6. Strauss-defined LBBB was independently evaluated by three physicians and an automated algorithm.

We identified 196 QRS markers on unfiltered ECGs. Strong filtering reduced the total marker count by 22% (153 vs. 196) and altered marker morphology, with 48% of notches converting to slurs and 38% of slurs disappearing. In contrast, moderate and mild filtering caused minimal changes, with ≤2% reduction in total markers and <10% of transformations. Under strong filtering, three physicians and an automated algorithm underdiagnosed Strauss-LBBB in about one-third to one-half of cases. Diagnostic performance remained comparable to unfiltered ECGs at 0.05–60 Hz for physicians and at 0.05–150 Hz for the algorithm.

ECG signal filtering significantly influences the ability of physicians and an automated algorithm to diagnose Strauss-LBBB. To correctly diagnose Strauss-LBBB, a bandwidth of at least 0.05–60 Hz is required for physicians and 0.05–150 Hz for the automated algorithm.

QRS marker transformations