Background: Left bundle branch area pacing (LBBAP) has recently been introduced as a novel physiological pacing strategy. Within LBBAP, distinction is made between left bundle branch pacing (LBBP) and left ventricular septal pacing (LVSP, no left bundle capture). Objective: To investigate acute electrophysiological effects of LBBP and LVSP as compared to intrinsic ventricular conduction. Methods: Fifty patients with normal cardiac function and pacemaker indication for bradycardia underwent LBBAP. Electrocardiography (ECG) characteristics were evaluated during pacing at various depths within the septum: starting at the right ventricular (RV) side of the septum: the last position with QS morphology, the first position with r’ morphology, LVSP and—in patients where left bundle branch (LBB) capture was achieved—LBBP. From the ECG’s QRS duration and QRS morphology in lead V1, the stimulus- left ventricular activation time left ventricular activation time (LVAT) interval were measured. After conversion of the ECG into vectorcardiogram (VCG) (Kors conversion matrix), QRS area and QRS vector in transverse plane (Azimuth) were determined. Results: QRS area significantly decreased from 82 ± 29 µVs during RV septal pacing (RVSP) to 46 ± 12 µVs during LVSP. In the subgroup where LBB capture was achieved (n = 31), QRS area significantly decreased from 46 ± 17 µVs during LVSP to 38 ± 15 µVs during LBBP, while LVAT was not significantly different between LVSP and LBBP. In patients with normal ventricular activation and narrow QRS, QRS area during LBBP was not significantly different from that during intrinsic activation (37 ± 16 vs. 35 ± 19 µVs, respectively). The Azimuth significantly changed from RVSP (−46 ± 33°) to LVSP (19 ± 16°) and LBBP (−22 ± 14°). The Azimuth during both LVSP and LBBP were not significantly different from normal ventricular activation. QRS area and LVAT correlated moderately (Spearman’s R = 0.58). Conclusions: ECG and VCG indices demonstrate that both LVSP and LBBP improve ventricular dyssynchrony considerably as compared to RVSP, to values close to normal ventricular activation. LBBP seems to result in a small, but significant, improvement in ventricular synchrony as compared to LVSP.

Left bundle branch block (LBBB) morphology is associated with improved outcome of cardiac resynchronisation therapy (CRT) and is an important criterion for patient selection. There are, however, multiple definitions for LBBB. Moreover, applying these definitions seems subjective. We investigated the inter- and intraobserver agreement in the determination of LBBB using available definitions, and clinicians' judgement of LBBB. Observers were provided with 12‑lead ECGs of 100 randomly selected CRT patients. Four observers judged the ECGs based on different LBBB-definitions (ESC, AHA/ACC/HRS, MADIT, and Strauss). Additionally, four implanting cardiologists scored the same 100 ECGs based on their clinical judgement. Observer agreement was summarized through the proportion of agreement (P) and kappa coefficient (k). Relative intra-observer agreement using different LBBB definitions, and within clinical judgement was moderate (range k 0.47-0.74 and k = 0.76 (0.14), respectively). The inter-observer agreement between observers using LBBB definitions as well as between clinical observers was minimal to weak (range k 0.19-0.44 and k = 0.35 (0.20), respectively). The probability of classifying an ECG as LBBB by available definitions varied considerably (range 0.20-0.76). The agreement between different definitions of LBBB ranged from good (P = 0.95 (0.07)) to weak (P = 0.40 (0.22)). Furthermore, correlation between the different LBBB definitions and clinical judgement was poor (range phi 0.30-0.55). Significant variation in the probability of classifying LBBB is present in using different definitions and clinical judgement. Considerable intra- and inter-observer variability adds to this variation. Interdefinition agreement varies significantly and correlation of clinical judgement with LBBB classification by definitions is modest at best.

This study aims to identify and visualize electrocardiogram (ECG) features using an explainable deep learning–based algorithm to predict cardiac resynchronization therapy (CRT) outcome. Its performance is compared with current guideline ECG criteria and QRS_AREA. A deep learning algorithm, trained on 1.1 million ECGs from 251 473 patients, was used to compress the median beat ECG, thereby summarizing most ECG features into only 21 explainable factors (FactorECG). Pre-implantation ECGs of 1306 CRT patients from three academic centres were converted into their respective FactorECG. FactorECG predicted the combined clinical endpoint of death, left ventricular assist device, or heart transplantation [c-statistic 0.69, 95% confidence interval (CI) 0.66–0.72], significantly outperforming QRS_AREA and guideline ECG criteria [c-statistic 0.61 (95% CI 0.58–0.64) and 0.57 (95% CI 0.54–0.60), P < 0.001 for both]. The addition of 13 clinical variables was of limited added value for the FactorECG model when compared with QRS_AREA (Δ c-statistic 0.03 vs. 0.10). FactorECG identified inferolateral T-wave inversion, smaller right precordial S- and T-wave amplitude, ventricular rate, and increased PR interval and P-wave duration to be important predictors for poor outcome. An online visualization tool was created to provide interactive visualizations (https://crt.ecgx.ai). Requiring only a standard 12-lead ECG, FactorECG held superior discriminative ability for the prediction of clinical outcome when compared with guideline criteria and QRS_AREA, without requiring additional clinical variables. End-to-end automated visualization of ECG features allows for an explainable algorithm, which may facilitate rapid uptake of this personalized decision-making tool in CRT. First, an artificial intelligence algorithm (variational auto-encoder) was pretrained on 1.1 million electrocardiograms (ECGs) to learn the underlying continuous factors that generate the ECG (i.e. the FactorECG). In this process, the variational auto-encoder (VAE) learns to reconstruct ECGs as accurately as possible using only 21 continuous factors without any human input. In the training phase, the pre-procedural median beat ECGs of 1306 cardiac resynchronization therapy (CRT) patients were each converted into their FactorECG. These 21 factors were subsequently used as input in a Cox model to predict the primary composite endpoint of left ventricular assist device implantation, heart transplantation and all-cause death, and the secondary endpoint of echocardiographic response. FactorECG significantly improved outcome prediction following CRT when compared with the current guidelines and QRS_AREA. The algorithm is explainable by using the decoder to visualize the effect of the ECG factors that significantly predicted outcome on the median beat ECG morphology. Here, for example, the influence of Factor 9 (F₉) is visualized, where higher values represent a more left bundle branch block-like ECG morphology and lower values, a more right bundle branch block-like morphology. CRT, cardiac resynchronization therapy; DNN, deep neural network; ECG, electrocardiogram; HTx, heart transplantation; LVAD, left ventricular assist device.

Pacing remote from the latest electrically activated site (LEAS) in the left ventricle (LV) may diminish response to cardiac resynchronization therapy (CRT). We tested whether proximity of LV pacing site (LVPS) to LEAS, determined by non-invasive three-dimensional electrical activation mapping [electrocardiographic Imaging (ECGI)], increased likelihood of CRT response. Consecutive CRT patients underwent ECGI and chest/heart computed tomography 6–24 months of post-implant. Latest electrically activated site and the distance to LVPS (d_p) were assessed. Left ventricular end-systolic volume (LVESV) reduction of ≥15% at clinical follow-up defined response. Logistic regression probabilistically modelled non-response; variables included demographics, heart failure classification, left bundle branch block (LBBB), ischaemic heart disease (IHD), atrial fibrillation, QRS duration, baseline ejection fraction (EF) and LVESV, comorbidities, use of CRT optimization algorithm, angiotensin-converting enzyme inhibitor(ACE)/angiotensin-receptor blocker (ARB), beta-blocker, diuretics, and d_p. Of 111 studied patients [64 ± 11 years, EF 28 ± 6%, implant duration 12 ± 5 months (mean ± SD), 98% had LBBB, 38% IHD], 67% responded at 10 ± 3 months post CRT-implant. Latest electrically activated sites were outside the mid-to-basal lateral segments in 35% of the patients. d_p was 42 ± 23 mm [31 ± 14 mm for responders vs. 63 ± 24 mm non-responders (P < 0.001)]. Longer d_p and the lack of use of CRT optimization algorithm were the only independent predictors of non-response [area under the curve (AUC) 0.906]. d_p of 47 mm delineated responders and non-responders (AUC 0.931). The distance between LV pacing site and latest electrical activation is a strong independent predictor for CRT response. Non-invasive electrical evaluation to characterize intrinsic activation and guide LV lead deployment may improve CRT efficacy.

Cardiac resynchronization therapy (CRT) is a well-known treatment modality for patients with a reduced left ventricular ejection fraction accompanied by a ventricular conduction delay. However, a large proportion of patients does not benefit from this therapy. Better patient selection may importantly reduce the number of non-responders. Here, we review the strengths and weaknesses of the electrocardiogram (ECG) markers currently being used in guidelines for patient selection, e.g., QRS duration and morphology. We shed light on the current knowledge on the underlying electrical substrate and the mechanism of action of CRT. Finally, we discuss potentially better ECG-based biomarkers for CRT candidate selection, of which the vectorcardiogram may have high potential.