Machine Learning Prediction of Cardiac Resynchronisation Therapy Response From Combination of Clinical and Model-Driven Data

Background: Up to 30–50% of chronic heart failure patients who underwent cardiac resynchronization therapy (CRT) do not respond to the treatment. Therefore, patient stratification for CRT and optimization of CRT device settings remain a challenge.Objective: The main goal of our study is to develop a predictive model of CRT outcome using a combination of clinical data recorded in patients before CRT and simulations of the response to biventricular (BiV) pacing in personalized computational models of the cardiac electrophysiology.Materials and Methods: Retrospective data from 57 patients who underwent CRT device implantation was utilized. Positive response to CRT was defined by a 10% increase in the left ventricular ejection fraction in a year after implantation. For each patient, an anatomical model of the heart and torso was reconstructed from MRI and CT images and tailored to ECG recorded in the participant. The models were used to compute ventricular activation time, ECG duration and electrical dyssynchrony indices during intrinsic rhythm and BiV pacing from the sites of implanted leads. For building a predictive model of CRT response, we used clinical data recorded before CRT device implantation together with model-derived biomarkers of ventricular excitation in the left bundle branch block mode of activation and under BiV stimulation. Several Machine Learning (ML) classifiers and feature selection algorithms were tested on the hybrid dataset, and the quality of predictors was assessed using the area under receiver operating curve (ROC AUC). The classifiers on the hybrid data were compared with ML models built on clinical data only.Results: The best ML classifier utilizing a hybrid set of clinical and model-driven data demonstrated ROC AUC of 0.82, an accuracy of 0.82, sensitivity of 0.85, and specificity of 0.78, improving quality over that of ML predictors built on clinical data from much larger datasets by more than 0.1. Distance from the LV pacing site to the post-infarction zone and ventricular activation characteristics under BiV pacing were shown as the most relevant model-driven features for CRT response classification.Conclusion: Our results suggest that combination of clinical and model-driven data increases the accuracy of classification models for CRT outcomes.

Scalable and reversible axonal neuromodulation of the sympathetic chain for cardiac control

Maladaptation of the sympathetic nervous system contributes to the progression of cardiovascular disease and risk for sudden cardiac death, the leading cause of mortality worldwide. Axonal modulation therapy (AMT) directed at the paravertebral chain blocks sympathetic efferent outflow to the heart, and may be a promising strategy to mitigate excess disease-associated sympathoexcitation. The present work evaluates AMT, directed at the sympathetic chain, in blocking sympathoexcitation using a porcine model. In anesthetized porcine (n=14), we applied AMT to the right T1-T2 paravertebral chain and performed electrical stimulation of the distal portion of the right sympathetic chain (RSS). RSS-evoked changes in heart rate, contractility, ventricular activation recovery interval (ARI), and norepinephrine release were examined with and without kilohertz frequency alternating current block (KHFAC). To evaluate efficacy of AMT in the setting of sympathectomy, evaluations were performed in the intact state and repeated after left and bilateral sympathectomy. We found strong correlations between AMT intensity and block of sympathetic stimulation-evoked changes in cardiac electrical and mechanical indices (r=0.83-0.96, effect size d=1.9-5.7), as well as evidence of sustainability and memory. AMT significantly reduced RSS-evoked left ventricular interstitial norepinephrine release, as well as coronary sinus norepinephrine levels. Moreover, AMT remained efficacious following removal of the left sympathetic chain, with similar mitigation of evoked cardiac changes and reduction of catecholamine release. With growth of neuromodulation, an on-demand or reactionary system for reversible AMT may have therapeutic potential for cardiovascular disease-associated sympathoexcitation.

Physiology and Practicality of Left Ventricular Septal Pacing

Left ventricular septal pacing (LVSP) and left bundle branch pacing (LBBP) have been introduced to maintain or correct interventricular and intraventricular (dys)synchrony. LVSP is hypothesised to produce a fairly physiological sequence of activation, since in the left ventricle (LV) the working myocardium is activated first at the LV endocardium in the low septal and anterior free-wall regions. Animal studies as well as patient studies have demonstrated that LV function is maintained during LVSP at levels comparable to sinus rhythm with normal conduction. Left ventricular activation is more synchronous during LBBP than LVSP, but LBBP produces a higher level of intraventricular dyssynchrony compared to LVSP. While LVSP is fairly straightforward to perform, targeting the left bundle branch area may be more challenging. Long-term effects of LVSP and LBBP are yet to be determined. This review focuses on the physiology and practicality of LVSP and provides a guide for permanent LVSP implantation.

Comparing Non-invasive Inverse Electrocardiography With Invasive Endocardial and Epicardial Electroanatomical Mapping During Sinus Rhythm

This study presents a novel non-invasive equivalent dipole layer (EDL) based inverse electrocardiography (iECG) technique which estimates both endocardial and epicardial ventricular activation sequences. We aimed to quantitatively compare our iECG approach with invasive electro-anatomical mapping (EAM) during sinus rhythm with the objective of enabling functional substrate imaging and sudden cardiac death risk stratification in patients with cardiomyopathy. Thirteen patients (77% males, 48 ± 20 years old) referred for endocardial and epicardial EAM underwent 67-electrode body surface potential mapping and CT imaging. The EDL-based iECG approach was improved by mimicking the effects of the His-Purkinje system on ventricular activation. EAM local activation timing (LAT) maps were compared with iECG-LAT maps using absolute differences and Pearson’s correlation coefficient, reported as mean ± standard deviation [95% confidence interval]. The correlation coefficient between iECG-LAT maps and EAM was 0.54 ± 0.19 [0.49–0.59] for epicardial activation, 0.50 ± 0.27 [0.41–0.58] for right ventricular endocardial activation and 0.44 ± 0.29 [0.32–0.56] for left ventricular endocardial activation. The absolute difference in timing between iECG maps and EAM was 17.4 ± 7.2 ms for epicardial maps, 19.5 ± 7.7 ms for right ventricular endocardial maps, 27.9 ± 8.7 ms for left ventricular endocardial maps. The absolute distance between right ventricular endocardial breakthrough sites was 30 ± 16 mm and 31 ± 17 mm for the left ventricle. The absolute distance for latest epicardial activation was median 12.8 [IQR: 2.9–29.3] mm. This first in-human quantitative comparison of iECG and invasive LAT-maps on both the endocardial and epicardial surface during sinus rhythm showed improved agreement, although with considerable absolute difference and moderate correlation coefficient. Non-invasive iECG requires further refinements to facilitate clinical implementation and risk stratification.

Efficacy of left bundle branch area pacing in patients with indication for cardiac resynchronization therapy

Background Cardiac resynchronization therapy (CRT) with biventricular pacing has demonstrated clinical benefits in heart failure patients with left bundle branch block (LBBB) and ventricular dysfunction. Left bundle branch area pacing (LBBAP) results in a relatively short QRS duration (QRSd) with fast left ventricular activation and could be considered as an alternative to conventional CRT. Purpose The aim of the present study was to evaluate the feasibility and outcomes of LBBAP in patients with indications for CRT. Methods Consecutive patients with indications for CRT were included. LBBAP was performed via transventricular septal approach (1–3). We aimed to achieve a paced QRS with right bundle branch conduction delay morphology, a stimulus to peak left ventricular activation time (S-LVAT) <100ms and/or a QRSd ≤130ms. AV delay programming was individualized in patients in sinus rhythm, taking consideration of the AV conduction, programming the one that generated the shortest QRSd at rest. Rate adaptive AV was also activated in these patients. Pacing electrical and echocardiographic parameters were recorded at baseline and during follow-up. Results LBBAP was achieved in 19 of 21 (90.5%) patients with indication for CRT. Indications were heart failure with LBBB and left ventricular ejection fraction (LVEF) ≤35% in 8 (42%), AV node ablation or AV block with LVEF <50% and high expected RV pacing burden in 9 (47%), 1 pacing-induced cardiomyopathy and 1 patient with biventricular pacemaker malfunction (high LV capture threshold). The mean follow-up was 4.6±1.7 months and the percentage of ventricular pacing was 93.4±13.9%. There were no device-related complications during this period. LBBA capture threshold was 0.6±0.3V at 0.4ms at the implantation, and remained stable (0.7±0.1 V, p=0.17). The lead impedance and R-wave amplitude at implantation were 636±106 ohms and 13.4±6.8 mV, and 541±88 ohms and 13.0±5.1 mV during the follow-up (p<0.001 and p=0.27, respectively). Mean S-LVAT was 85.5±13.9 ms, and mean QRSd was 122±9 ms, that remained stable during follow-up (122 vs 124 ms, p=0.21). In patients with LBBB, a significant narrowing of paced QRSd was achieved (160.9±16.7 vs. 123.9±9.7 ms, p<0.001). Mean LVEF increased by 15.9%, from 35.4±8.9% at baseline to 51.3±9.8% at follow-up (p<0.001) in the overall population, and 14.5% (from 32.7±4.8% to 47.2±10.7%, p=0.001) in patients with LBBB. After one month, estimated time for elective replacement was 11.9±0.4 years. Conclusions LBBPA was successfully achieved in 90.5% of the patients with indication for CRT, with good and stable pacing electrical parameters, long estimated battery longevity and relatively narrow QRS, and was associated with improvement in cardiac function. LBBAP may be considered as a first-line option for patients with indications for CRT. FUNDunding Acknowledgement Type of funding sources: None.

Prospective evaluation of the learning curve and electrical characteristics of left bundle branch area pacing

Background Left bundle branch area pacing (LBBAP) has recently been introduced as a physiological pacing technique with a synchronous ventricular activation. Objective To prospectively evaluate the feasibility and learning curve, as well as the electrical characteristics of LBBAP. Methods In 80 consecutive LBBAP pacemaker patients, ECG characteristics during intrinsic rhythm, RV septum pacing (RVSP) and LBBAP were evaluated. From the ECG's QRS duration and LVAT (stimulus to V6 R-wave peak time, RWPT) were measured. Also, the left bundle branch potential (LBBpot) to V6 RWPT interval was measured and compared to the LVAT. After conversion of the ECG into VCG (Kors conversion matrix), QRS area, as measurement for electrical dyssynchrony, was calculated. Results Permanent lead implantation was successful in 77/80 patients (96%) undergoing an attempt at LBBAP. LBBAP lead implantation time as well as fluoroscopy time were significantly shorter during last 25% of implantation compared to first 25% of implantations (17±5 min vs. 33±16 min and 12±7 min vs. 21±13 min, respectively, panel A and B). LBB capture was obtained in 54/80 patients (68%). In 36/45 patients (80%) with intact AV conduction and narrow QRS an LBBpot was present. The mean interval between the LBBpot and the onset of QRS was 22±6 ms. In the patients with narrow QRS (n=45), QRS duration increased significantly during both RVSP (139±24 ms) and LBBAP (123±21 ms), compared to intrinsic rhythm (95±13 ms). QRS area on the other hand, increased during both RVSP (73±20 μVs) but decreased during LBBAP (41±15 μVs), to values close to intrinsic rhythm (32±16 μVs, panel C). For all patients, QRS area was significantly lower in patients with LBB capture compared to patients without capture (43±18 μVs vs 54±21 μVs, respectively). In patients with LBB capture (n=54), LVAT was significantly shorter compared to patients without LBB capture (75±14 vs. 88±9 ms, respectively). In the patients with LBB capture, there was a significant correlation between the LBBpot – V6 RWPT and S – V6 RWPT intervals (Pearson correlation 0.739, P<0.001). Conclusion LBBAP is a safe and feasible technique, with a clear learning curve that seems obtained after ± 40–60 implantations. LBB capture is obtained in two-thirds of patients. Although QRS duration remains prolonged, LBBAP largely restores ventricular electrical synchrony to values close to intrinsic (narrow QRS) rhythm. FUNDunding Acknowledgement Type of funding sources: None.

Machine Learning prediction of cardiac resynchronisation therapy response from combination of clinical and model-driven data

BackgroundUp to 30%-50% of chronic heart failure patients who underwent cardiac resynchronization therapy (CRT) do not respond to the treatment. Therefore, patient stratification for CRT and optimization of CRT device settings remain a challenge.ObjectiveThe main goal of our study is to develop a predictive model of CRT outcome using a combination of clinical data recorded in patients before CRT and simulations of the response to biventricular (BiV) pacing in personalized computational models of the cardiac electrophysiology.Materials and MethodsRetrospective data from 57 patients who underwent CRT device implantation was utilized. Positive response to CRT was defined by a 10% increase in the left ventricular ejection fraction in a year after implantation. For each patient, an anatomical model of the heart and torso was reconstructed from MRI and CT images and tailored to ECG recorded in the participant. The models were used to compute ventricular activation time, ECG duration and electrical dyssynchrony indices during intrinsic rhythm and BiV pacing from active poles of leads. For building a predictive model of CRT response, we used clinical data recorded before CRT device implantation together with model-derived biomarkers of ventricular excitation in the left bundle branch block mode of activation and under BiV stimulation. Several Machine Learning (ML) classifiers and feature selection algorithms were tested on the hybrid dataset, and the quality of predictors was assessed using the area under receiver operating curve (ROC AUC). The classifiers on the hybrid data were compared with ML models built on clinical data only.ResultsThe best ML classifier utilizing a hybrid set of clinical and model-driven data demonstrated ROC AUC of 0.82, an accuracy of 0.82, sensitivity of 0.85, and specificity of 0.78, improving quality over that of ML predictors built on clinical data from much larger datasets. Distance from the LV pacing site to the post-infarction zone and ventricular activation characteristics under BiV pacing were shown as the most relevant model-driven features for CRT response classification.ConclusionOur results suggest that combination of clinical and model-driven data increases the accuracy of classification models for CRT outcomes.