Paul J. Wang:

Welcome to the monthly podcast, On the BEAT, for Circulation: Arrhythmia and Electrophysiology. I'm Dr. Paul Wang, Editor in Chief, with some of the key highlights from this month's issue.

Albert Feeny and Associates used unsupervised machine learning of electrocardiogram [ECG] waveforms to identify cardiac resynchronization therapy [CRT] subgroups to differentiate outcomes beyond QRS duration and left bundle branch block. They retrospectively analyzed 946 CRT patients with conduction delay. Principal component analysis [PCA] dimensionality reduction obtained a 2-dimensional representation of pre-CRT 12-lead QRS waveforms. K-means clustering of the 2-dimensional PCA representation of 12-lead QRS waveforms identified two patient subgroups [QRS PCA groups]. Vectorcardiographic QRS area was also calculated. They examined two primary outcomes: (1) composite endpoint of death, left ventricular assist device, or heart transplant, and (2) degree of echocardiographic left ventricular ejection fraction [LVEF] change after CRT. Compared to QRS PCA group 2 (n = 425), Group 1 (n=521) had a lower risk for achieving the composite endpoint (hazard ratio of 0.44, P < 0.001) and experienced greater mean LVEF improvement (11.1% versus 4.8%, P < 0.001), even among left bundle branch block patients with QRS duration, 150 milliseconds or greater (hazard ratio 0.45, P < 0.001; mean LVF change 12.5% versus 7.3%, P=0.001). A stratification scheme combining QRS area and QRS PCA group identified left bundle branch block patients with similar outcomes as non left bundle branch block patients (hazard ratio 1.32, mean difference LVEF change 0.8%). That stratification scheme also identified left bundle branch block patients with QRS duration less than 150 ms is comparable outcomes to left bundle branch patients with QRS duration 150 ms or greater (hazard ratio 0.93, mean difference in LVF change -0.2%). The authors concluded that unsupervised machine learning of ECG waveforms identified CRT subgroups with relevance beyond left bundle branch block and QRS duration.

In our next paper, Julie Shade, Rheeda Ali and Associates combined machine learning [ML] and personalized computational modeling to predict, prior to pulmonary vein isolation [PVI], which patients are most likely to experience atrial fibrillation [AF] recurrence after PVI. The single center retrospective proof of concept study included 32 patients with documented paroxysmal AF who underwent PVI and had pre-procedural late gadolinium enhanced magnetic resonance imaging [LGE MRI]. For each patient, a personalized computational model of the left atrium simulated AF induction via rapid pacing features were derived from pre-PVI LG MRI images and from results of simulations [SIM] AF. The most predictive features used to input to a quadratic discrimination analysis ML classifier, which was trained, optimized, and evaluated with a 10-fold nested cross validation to predict the probability of AF recurrence post PVI. In the cohort, the ML classifier predicted probability of AF recurrence with an average validation, sensitivity, and specificity of 82% and 89% respectively, and a validation AUC of 0.82. Dissecting the relative contributions of simulations SIM AF and raw images to the predictive capability of the ML classifier, they found that only when features from simulation SIM AF were used to train the ML classifier, its performance retained similar (validation AUC equals 0.81). However, when only features classified from raw images were used for training, the validation AUC significantly decreased (0.47).

In our next paper, Sarah Vermij and Associates examined sodium channel NaV 1.5 localization and function mutations in the gene and coding the sodium channel NaV 1.5 caused various cardiac arrhythmias. The authors use novel single-molecule localization [S-M-L-M] and computational modeling to define nanoscale features of NaV 1.5 localization and distribution at the lateral membrane [L-M], the LM groove, and T-tubules in cardiomyocytes from wild-type (N=3), dystrophin-deficient (mdx; N=3) mice, and mice expressing C-terminally truncated NaV 1.5 (ΔSIV; N=3). The authors assessed T-tubules sodium current by recording whole-cell sodium currents in control (N=5) in detubulated (N=5) wild-type cardiomyocytes. The authors found that NaV 1.5 organizes as distinct clusters in the groove and T-tubules which density, distribution, and organization partially depend on SIV and dystrophin. They found that overall reduction in NaV 1.5 expression expressed in mdx and ΔSIV cells result in a non-uniform distribution with NaV 1.5 being specifically reduced at the groove ΔSIV and increased in T-tubules of mdx cardiomyocytes. A T-tubules sodium current could, however, not be demonstrated. The authors concluded that NaV 1.5 mutations may site-specifically affect NaV 1.5 localization and distribution at the lateral membrane and T-tubules, depending on site-specific interacting proteins.

In our next paper, Sharan Sharma, Mohit Turagam, and associates studied strategies to improve patient comfort related to pericardial access. They conducted a multi-centered retrospective study, including 104 patients who underwent epicardial ventricular tachycardia [VT] ablation and Lariat left atrial appendage occlusion. They compared 53 patients who received post-procedural intrapericardial liposomal bupivacaine (LB)+oral colchicine (LB group) and 51 patients who received colchicine alone (non-LB group). Lyposomal bupivacaine was associated with significant lowering of median pain scale at 6 hours (1.0 versus 8.0, P<0.001), at 12 hours (1.0 versus 6.0, P<0.001), and up to 48 hours post-procedure. Incidence of acute severe pericarditis delayed pericardial effusion and gastrointestinal adverse effects were similar in both groups. Median length of stay was significantly lower in the lyposomal bupivacaine pain group (2.0 versus 3.0, P<0.001). Subgroup analysis demonstrated similar favorable outcomes in both Lariat and epicardial VT ablation groups.

In our next paper, Sergio Callegari, Emilio Macchi, and Associates characterize the fibrosis (amount, architecture, cellular components, and ultrastructure) in left atrial biopsies from 121 patients with persistent/long-lasting atrial fibrillation [AF] (group 1; 59 males; 60 years of age; 91 mitral disease-related AF, 30 nonmitral disease-related AF) and 39 patients in sinus rhythm with mitral valve regurgitation (group 2; 32 males; 59 years of age). 10 autopsy hearts served as controls. Qualitatively, the fibrosis exhibited the same characteristics in all cases and displayed particular architectural scenarios (which the authors arbitrarily divided into four stages) ranging from isolated foci to confluent sclerotic areas. The percentage of fibrosis was larger and in a more advanced stage in group 1 versus group 2 and within group 1, in patients with rheumatic disease versus non-rheumatic cases. In AF patients with mitral disease and no rheumatic disease, the percentage of fibrosis and the fibrosis stages correlated with both left atrial volume index and AF duration. The fibrotic areas mainly consisted of type I collagen with only a minor cellular component (especially fibroblasts/myofibroblasts; average value range 69–150 cells/mm2, depending on the areas in AF biopsies). A few fibrocytes, circulating and bone marrow-derived mesenchymal cells, were also detectable. The fibrosis-entrapped cardiomyocytes showed sarcolemmal damage and connexin 43 redistribution/internalization.

In our next paper, Shijie Zhou and Associates tested an automated localization system to identify the site of origin of left ventricular [LV] activation in real time using 12-lead ECG. The automated site of origin, solo system, consists of three steps: (1) localization of ventricular segment based on population templates, (2) population-based localization within a segment, and, (3) patient-specific site localization. Localization error was assessed by the distance between the known reference site and the estimated site. In 19 patients undergoing 21 catheter ablation procedures of scar related VT, solo accuracy was estimated using 552 LV left endocardial pacing sites pooled together and 25 VT-exit sites identified by contact mapping. For the 25 VT-exit sites, localization error of the population-based localization steps was within 10 mm. Patient-specific site localization achieved accuracy of within 3.5 mm after including up to 11 pacing (training) sites. Using 3 remotes (67.8 mm from the reference VT-exit site), and then 5 close pacing sites, resulted in localization error of 7.2 mm for the 25 identified VT-exit sites. In 2 emulated clinical procedure with 2 induced VT's, the solo system achieved accuracy within 4 mm.

In our next paper, Ryan Koene and associates examined outcomes of use of dofetilide in atrial fibrillation [AF] patients with left ventricular ejection fraction [LVEF]≤35% without prior implantable cardioverter defibrillator [ICD] cardiac resynchronization therapy [CRT], or AF ablation. An analysis of 168 consecutive patients from 2007 to 2016 was performed. Incidences of adverse events, drug discontinuation, ICD, or CRT implementation, LVEF improvement (>35%) and recovery (50% or greater), AF recurrence, and AF ablation were determined. Multi-variate regression analysis to identify predictors of LVEF improvement/recovery was performed. The mean age was 64 years. Dofetilide was discontinued prior to hospital discharge in 46 (27%) because of QT prolongation (14%), torsades de pointe or polymorphic ventricular tachycardia/ventricular fibrillation (6% [sustained 3%, nonsustained 3%]), in effectiveness (5%), and other causes (3%). At one year 43% remained on dofetilide. Freedom from AF was 42% at 1 year and 40% underwent future AF ablation. LVEF recovered to 50% or greater in 45% and an improved to greater than 35% in 73%. Predictor