Repolarization in bundle branch block, can artificial intelligence offer insights?

Abstract Background QT interval interpretation remains challenging in patients with bundle branch block (BBB), where QRS widening prolongs the QT interval due to delayed ventricular depolarization. A validated correction formula, QTm = QTBBB − (50% × QRSBBB), compensates for prolongation of depolarization time due to BBB, , enabling estimation of true repolarization time. However, its clinical impact on automated ECG prioritization systems requires validation. Objective To evaluate the effect of BBB-specific QT correction on automated severity classification and validate algorithm performance in detecting BBB and measuring intervals. Methods We analysed 250 ECGs selected from a proprietary database: 50 BBB cases (25 LBBB, 25 RBBB) and 200 controls (125 with normal conduction, 75 with non-specific QRS widening ≥110 ms). QTBBB and QRSBBB were defined as the longest globally measured QT and QRS intervals on the 12-lead ECG. The PulseAI neural network classified BBB and measured intervals. QT correction (QTm) was applied to BBB cases, followed by Fridericia’s rate adjustment. Severity classifications included: Emergent (≥500 ms), Significant (≥470 ms), Borderline (≥430 ms), and Normal (<430 ms). Results The final analysis included 246 ECGs (4 excluded for noise), with 49 patients exhibiting either left or right BBB. The PulseAI neural network demonstrated high accuracy in detecting BBB, achieving a sensitivity of 92% (45/49) with a positive predictive value of 100%. Agreement between PulseAI and reference measurements was strong, with mean differences of 2.4 ± 10.5 ms for QT intervals and 1.7 ± 7.4 ms for QRS durations. Application of the BBB-specific QT correction formula resulted in substantial reclassification: all 4 patients initially classified as emergent and all 15 initially classified as significant were reclassified to lower-risk groups, leaving zero in these categories after correction. Borderline classifications decreased from 24 to 4, while the normal category expanded from 6 to 45 patients. These findings demonstrate that the correction formula effectively reduces overestimation of QT prolongation severity in patients with BBB, by dissociating the contribution of prolonged depolarization from true repolarization time. Conclusion BBB-specific QT correction eliminated emergent and significant classifications in all cases, reclassifying 88% of BBB patients to lower-risk categories. High algorithm accuracy supports reliable integration into automated systems, mitigating overestimation of repolarization risk in BBB and reducing unnecessary clinical alerts. This approach improves risk stratification by prioritizing true repolarization abnormalities rather than changes due to prolonged depolarization. [1] Bogossian, H., Linz, D., Heijman, J., Bimpong-Buta, N.Y., Bandorski, D., Frommeyer, G., Erkapic, D., Seyfarth, M., Zarse, M. and Crijns, H.J., 2020. QTc evaluation in patients with bundle branch block. IJC Heart & Vasculature, 30, p.100636.