Characterization of cardiac resynchronization therapy response through machine learning and personalized models

INTRODUCTION

The characterization and selection of heart failure (HF) patients for cardiac resynchronization therapy (CRT) remain challenging, with around 30% non-responder rate despite following current guidelines. This study aims to propose a novel hybrid approach, integrating machine-learning and personalized models, to identify explainable phenogroups of HF patients and predict their CRT response.

METHODS

The paper proposes the creation of a complete personalized model population based on preoperative CRT patient strain curves. Based on the parameters and features extracted from these personalized models, phenotypes of patients are identified thanks to a clustering algorithm and a random forest classification is provided.

RESULTS

A close match was observed between the 162 experimental and simulated myocardial strain curves, with a mean RMSE of 4.48% (±1.08) for the 162 patients. Five phenogroups of personalized models were identified from the clustering, with response rates ranging from 52% to 94%. The classification results show a mean area under the curves (AUC) of 0.86 ± 0.06 and provided a feature importance analysis with 22 features selected. Results show both regional myocardial contractility (from 22.5% to 33.0%), tissue viability and electrical activation delays importance on CRT response for each HF patient (from 55.8 ms to 88.4 ms).

DISCUSSION

The patient-specific model parameters' analysis provides an explainable interpretation of HF patient phenogroups in relation to physiological mechanisms that seem predictive of the CRT response. These novel combined approaches appear as promising tools to improve understanding of LV mechanical dyssynchrony for HF patient characterization and CRT selection.

Myocardial strain of the left ventricle by speckle tracking echocardiography: From physics to clinical practice

Speckle tracking echocardiography (STE) is a reliable imaging technique of recognized clinical value in several settings. This method uses the motion of ultrasound backscatter speckles within echocardiographic images to derive myocardial velocities and deformation parameters, providing crucial insights on several cardiac pathological and physiological processes. Its feasibility, reproducibility, and accuracy have been widely demonstrated, being myocardial strain of the various chambers inserted in diagnostic algorithms and guidelines for various pathologies. The most important parameters are Global longitudinal strain (GLS), Left atrium (LA) reservoir strain, and Global Work Index (GWI): based on large studies the average of the lower limit of normality are -16%, 23%, and 1442 mmHg%, respectively. For GWI, it should be pointed out that myocardial work relies primarily on non-invasive measurements of blood pressure and segmental strain, both of which exhibit high variability, and thus, this variability constitutes a significant limitation of this parameter. In this review, we describe the principal aspects of the theory behind the use of myocardial strain, from cardiac mechanics to image acquisition techniques, outlining its limitation, and its principal clinical applications: in particular, GLS have a role in determine subclinical myocardial dysfunction (in cardiomyopathies, cardiotoxicity, target organ damage in ambulatory patients with arterial hypertension) and LA strain in determine the risk of AF, specifically in ambulatory patients with arterial hypertension.

The value of left ventricular strain-volume loops in predicting response to cardiac resynchronization therapy

BACKGROUND

Three-dimensional (3D) speckle tracking imaging (STI) allows the simultaneous assessment of left ventricular (LV) strain and volume. We aim to explore the value of LV strain-volume loops in predicting response to cardiac resynchronization therapy (CRT).

METHODS

Forty heart failure (HF) patients scheduled for CRT and twenty healthy individuals were enrolled. All subjects received a 3D echocardiography and 3D STI analysis to acquire LV global and segmental principal strain (PS) and volume simultaneously. Values were plotted in a Cartesian system to construct PS-volume loop which was assessed using the two characteristics of the linear fitting curve: the slope and the coefficient of determination (R²-S/D coupling).

RESULTS

HF patients at baseline showed significantly lower slope and R²-S/D coupling of all PS-volume loops than healthy subjects. As for as comparing Segmental PS-Global volume loop at baseline, Midseptal R²-S/D coupling was lower and Midlateral slope was higher in CRT responders than in non-responders. For each individual, the abnormal segmental heterogeneity of Midseptal slope and R²-S/D coupling were lower than Midlateral was observed only in responders. At follow-up, significant improvements of the Midseptal slope and R²-S/D coupling were observed in responders. Midseptal R²-S/D coupling at baseline was an independent predictor of CRT response and the cut-off value of 0.55 was recommended with sensitivity of 89% and specificity of 77%.

CONCLUSIONS

Analysis of strain-volume loops could provide unique information for predicting response to CRT. Assessment of septal myocardial wasted work at baseline is helpful to improve patient selection for CRT.

Cardiovascular magnetic resonance left ventricular strain in end-stage renal disease patients after kidney transplantation

BACKGROUND

Cardiovascular disease is a significant cause of morbidity and mortality in patients with end-stage renal disease (ESRD) and kidney transplant (KT) patients. Compared with left ventricular (LV) ejection fraction (LVEF), LV strain has emerged as an important marker of LV function as it is less load dependent. We sought to evaluate changes in LV strain using cardiovascular magnetic resonance imaging (CMR) in ESRD patients who received KT, to determine whether KT may improve LV function.

METHODS

We conducted a prospective multi-centre longitudinal study of 79 ESRD patients (40 on dialysis, 39 underwent KT). CMR was performed at baseline and at 12 months after KT.

RESULTS

Among 79 participants (mean age 55 years; 30% women), KT patients had significant improvement in global circumferential strain (GCS) (p = 0.007) and global radial strain (GRS) (p = 0.003), but a decline in global longitudinal strain (GLS) over 12 months (p = 0.026), while no significant change in any LV strain was observed in the ongoing dialysis group. For KT patients, the improvement in LV strain paralleled improvement in LVEF (57.4 ± 6.4% at baseline, 60.6% ± 6.9% at 12 months; p = 0.001). For entire cohort, over 12 months, change in LVEF was significantly correlated with change in GCS (Spearman's r = - 0.42, p < 0.001), GRS (Spearman's r = 0.64, p < 0.001), and GLS (Spearman's r = - 0.34, p = 0.002). Improvements in GCS and GRS over 12 months were significantly correlated with reductions in LV end-diastolic volume index and LV end-systolic volume index (all p < 0.05), but not with change in blood pressure (all p > 0.10).

CONCLUSIONS

Compared with continuation of dialysis, KT was associated with significant improvements in LV strain metrics of GCS and GRS after 12 months, which did not correlate with blood pressure change. This supports the notion that KT has favorable effects on LV function beyond volume and blood pessure control. Larger studies with longer follow-up are needed to confirm these findings.

Left ventricular regional remodeling and lead position during cardiac resynchronization therapy

BACKGROUND

Cardiac resynchronization therapy (CRT) induces segmental left ventricular (LV) remodeling. The LV lead position (LV-LP) affects response to CRT and remodeling.

OBJECTIVE

We aimed to assess segmental remodeling concordant, adjacent, and remote to LV-LP using cardiac computed tomography (CT).

METHODS

We included patients from the Empiric Versus Imaging-Guided Left Ventricular Lead Placement in Cardiac Resynchronization Therapy trial. Dynamic cardiac CT was performed at baseline and after 6 months. We assessed systolic wall thickening (WT) and exact LV-LP from the CT scans according to a 16-segment model. Response to CRT was defined as ≥15% reduction in LV end-systolic volume.

RESULTS

A total of 107 consecutive patients were included. The change in WT from baseline to follow-up was -19% (95% confidence interval [CI] -25% to -13%; P < .001) in concordant segments, -0.1% (95% CI -5% to 5%; P = .97) in adjacent segments, and 20% (95% CI -17% to 23%; P < .001) in remote segments. Diastolic wall thickness changed only marginally. Twenty patients (19%) were nonresponders at follow-up. In nonresponders with nonischemic cardiomyopathy, we observed a significant reduction in WT in concordant and adjacent segments with no increase in WT in remote segments.

CONCLUSION

During CRT, systolic WT increases in segments remote to LV-LP, decreases in concordant segments, and remains unchanged in adjacent segments. Only marginal changes occur in wall thickness. In nonresponders with nonischemic cardiomyopathy, deleterious changes in segmental myocardial function occur, and further studies on how to treat these patients best are warranted.

Imaging the Propagation of the Electromechanical Wave in Heart Failure Patients with Cardiac Resynchronization Therapy

BACKGROUND

Current electrocardiographic and echocardiographic measurements in heart failure (HF) do not take into account the complex interplay between electrical activation and local wall motion. The utilization of novel technologies to better characterize cardiac electromechanical behavior may lead to improved response rates with cardiac resynchronization therapy (CRT). Electromechanical wave imaging (EWI) is a noninvasive ultrasound-based technique that uses the transient deformations of the myocardium to track the intrinsic EW that precedes myocardial contraction. In this paper, we investigate the performance and reproducibility of EWI in the assessment of HF patients and CRT.

METHODS

EWI acquisitions were obtained in five healthy controls and 16 HF patients with and without CRT pacing. Responders (n = 8) and nonresponders (n = 8) to CRT were identified retrospectively on the basis of left ventricular (LV) reverse remodeling. Electromechanical activation maps were obtained in all patients and used to compute a quantitative parameter describing the mean LV lateral wall activation time (LWAT).

RESULTS

Mean LWAT was increased by 52.1 ms in HF patients in native rhythm compared to controls (P < 0.01). For all HF patients, CRT pacing initiated a different electromechanical activation sequence. Responders exhibited a 56.4-ms ± 28.9-ms reduction in LWAT with CRT pacing (P < 0.01), while nonresponders showed no significant change.

CONCLUSION

In this initial feasibility study, EWI was capable of characterizing local cardiac electromechanical behavior as it pertains to HF and CRT response. Activation sequences obtained with EWI allow for quantification of LV lateral wall electromechanical activation, thus providing a novel method for CRT assessment.

Left ventricular markers of mortality and ventricular arrhythmias in heart failure patients with cardiac resynchronization therapy

Aims

Cardiac resynchronization therapy (CRT) reduces morbidity and mortality in heart failure. However, prediction of the outcome remains difficult. We aimed to investigate for echocardiographic predictors of ventricular arrhythmias and fatal outcome and to explore how myocardial function is changed by biventricular pacing in heart failure.

Methods and results

We prospectively included 170 heart failure patients (66 ± 10 years, New York Heart Association class 2.8 ± 0.5, 48% ischaemic cardiomyopathy) and recorded ventricular arrhythmias and fatal end point defined as death, heart transplantation, or left ventricular assist device implantation during 2 years. Two-dimensional echocardiography was performed before and 6 months after CRT implantation. CRT response was defined as ≥15% reduction in end-systolic volume at 6 months. Speckle-tracking technique was performed to assess longitudinal and circumferential left ventricular function, defined as global longitudinal (GLS) and circumferential strain (GCS), and to assess mechanical dyssynchrony, defined as mechanical dispersion. GLS before CRT was a predictor of fatal end point independently of CRT response [hazard ratio, HR 1.14 (1.02–1.27), P = 0.02]. Patients with GLS better than −8.3% showed event-free survival benefit (log rank, P < 0.001). Mechanical dispersion at 6 months was an independent predictor of ventricular arrhythmias [HR 1.20 (1.06–1.35), P = 0.005]. CRT responders (59%) had improvement of both GLS and GCS.

Conclusion

In heart failure patients with CRT, worse longitudinal function before CRT was an important predictor of fatal outcome during 2 years, independently of CRT response. Mechanical dispersion at 6 months was a strong predictor of ventricular arrhythmias. CRT response by reverse remodelling was dependent on improvement of both longitudinal and circumferential function.

Comparison of septal strain patterns in dyssynchronous heart failure between speckle tracking echocardiography vendor systems

AIM

To analyze inter-vendor differences of speckle tracking echocardiography (STE) in imaging cardiac deformation in patients with dyssynchronous heart failure.

METHODS AND RESULTS

Eleven patients (all with LBBB, median age 60.7 years, 9 males) with implanted cardiac resynchronization therapy devices were prospectively included. Ultrasound systems of two vendors (i.e. General Electric and Philips) were used to record images in the apical four chamber view. Regional longitudinal strain patterns were analyzed with vendor specific software in the basal, mid and apical septal segments. Systolic strain (SS), time to peak strain (TTP) and septal rebound stretch (SRS) were determined during four pacing settings, resulting in 44 unique strain patterns per segment (total 132 patterns). Cross correlation was used to analyze the comparability of the shape of 132 normalized strain patterns. Correlation of strain patterns of the two systems was high (R(2) median: 0.68, interquartile range: 0.53-0.82). Accordingly, strain patterns of intrinsic rhythm were recognized equally using both systems, when divided into three types. GE based SS (18.9 ± 4.7%) was significantly higher than SS determined by the Philips system (13.4 ± 4.3%). TTP was slightly but non-significantly lower in GE (384 ± 77 ms) compared to Philips (404 ± 83 ms) derived strain signals. Correlation of SRS between the systems was poor, due to minor differences in the strain signal and timing of aortic valve closure.

CONCLUSIONS

The two systems provide similar shape of strain patterns. However, important differences are found in the amplitude, timing of systole and SRS. Until STE is standardized, clinical decision making should be restricted to pattern analysis.

Assessment of wasted myocardial work: a novel method to quantify energy loss due to uncoordinated left ventricular contractions

Left ventricular (LV) dyssynchrony reduces myocardial efficiency because work performed by one segment is wasted by stretching other segments. In the present study, we introduce a novel noninvasive clinical method that quantifies wasted energy as the ratio between work consumed during segmental lengthening (wasted work) divided by work during segmental shortening. The wasted work ratio (WWR) principle was studied in 6 anesthetized dogs with left bundle branch block (LBBB) and in 28 patients with cardiomyopathy, including 12 patients with LBBB and 10 patients with cardiac resynchronization therapy. Twenty healthy individuals served as controls. Myocardial strain was measured by speckle tracking echocardiography, and LV pressure (LVP) was measured by micromanometer and a previously validated noninvasive method. Segmental work was calculated by multiplying strain rate and LVP to get instantaneous power, which was integrated to give work as a function of time. A global WWR was also calculated. In dogs, WWR by estimated LVP and strain showed a strong correlation (r = 0.94) and good agreement with WWR by the LV micromanometer and myocardial segment length by sonomicrometry. In patients, noninvasive WWR showed a strong correlation (r = 0.96) and good agreement with WWR using the LV micromanometer. Global WWR was 0.09 ± 0.03 in healthy control subjects, 0.36 ± 0.16 in patients with LBBB, and 0.21 ± 0.09 in cardiomyopathy patients without LBBB. Cardiac resynchronization therapy reduced global WWR from 0.36 ± 0.16 to 0.17 ± 0.07 (P < 0.001). In conclusion, energy loss due to incoordinated contractions can be quantified noninvasively as the LV WWR. This method may be applied to evaluate the mechanical impact of dyssynchrony.

Multiview diffeomorphic registration: application to motion and strain estimation from 3D echocardiography

This paper presents a new registration framework for quantifying myocardial motion and strain from the combination of multiple 3D ultrasound (US) sequences. The originality of our approach lies in the estimation of the transformation directly from the input multiple views rather than from a single view or a reconstructed compounded sequence. This allows us to exploit all spatiotemporal information available in the input views avoiding occlusions and image fusion errors that could lead to some inconsistencies in the motion quantification result. We propose a multiview diffeomorphic registration strategy that enforces smoothness and consistency in the spatiotemporal domain by modeling the 4D velocity field continuously in space and time. This 4D continuous representation considers 3D US sequences as a whole, therefore allowing to robustly cope with variations in heart rate resulting in different number of images acquired per cardiac cycle for different views. This contributes to the robustness gained by solving for a single transformation from all input sequences. The similarity metric takes into account the physics of US images and uses a weighting scheme to balance the contribution of the different views. It includes a comparison both between consecutive images and between a reference and each of the following images. The strain tensor is computed locally using the spatial derivatives of the reconstructed displacement fields. Registration and strain accuracy were evaluated on synthetic 3D US sequences with known ground truth. Experiments were also conducted on multiview 3D datasets of 8 volunteers and 1 patient treated by cardiac resynchronization therapy. Strain curves obtained from our multiview approach were compared to the single-view case, as well as with other multiview approaches. For healthy cases, the inclusion of several views improved the consistency of the strain curves and reduced the number of segments where a non-physiological strain pattern was observed. For the patient, the improvement (pacing ON vs. OFF) in synchrony of regional strain correlated with clinician blind assessment and could be seen more clearly when using the multiview approach.

Electrical and mechanical ventricular activation during left bundle branch block and resynchronization

Cardiac resynchronization therapy (CRT) aims to treat selected heart failure patients suffering from conduction abnormalities with left bundle branch block (LBBB) as the culprit disease. LBBB remained largely underinvestigated until it became apparent that the amount of response to CRT was heterogeneous and that the therapy and underlying pathology were thus incompletely understood. In this review, current knowledge concerning activation in LBBB and during biventricular pacing will be explored and applied to current CRT practice, highlighting novel ways to better measure and treat the electrical substrate.

Simple regional strain pattern analysis to predict response to cardiac resynchronization therapy: rationale, initial results, and advantages

BACKGROUND

A classical strain pattern of early contraction in one wall and prestretching of the opposing wall followed by late contraction has previously been associated with left bundle branch block (LBBB) activation and short-term response to cardiac resynchronization therapy (CRT). Aims of this study were to establish the long-term predictive value of an LBBB-related strain pattern and to identify changes in contraction patterns during short-term and long-term CRT.

METHODS AND RESULTS

Sixty-seven patients with standard CRT criteria were prospectively enrolled between early 2009 and late 2010. Echocardiography including regional strain analysis by 2-dimensional speckle tracking was performed 1 week before implantation, at day 1, and 6 months after. Response was defined as a decrease in left ventricular end-systolic volume ≥ 15%. The predictive ability of a classical pattern was compared with time-to-peak measurements from velocity and deformation analysis. Forty-three patients (65%) were classified as responders. The presence of a classical pattern showed 91% specificity and 95% sensitivity for response and performed significantly better than time-to-peak parameters in prediction of response to CRT (P < .001, all). In responders, CRT acutely increased septal longitudinal peak systolic strain (-8.7% ± 3.6% to -11.1% ± 3%, P < .001) but not in nonresponders.

CONCLUSIONS

The classical pattern is highly predictive of response to CRT and superior to time-to-peak methods. Patients who obtain long-term reverse remodeling are characterized by short-term reversal of the classical strain pattern. These findings emphasize the value of recognizing potentially reversible strain patterns in selection of CRT candidates.

Mechano-energetics of the asynchronous and resynchronized heart

Abnormal electrical activation of the ventricles creates major abnormalities in cardiac mechanics. Local contraction patterns, as reflected by measurements of local strain, are not only out of phase, but often also show opposing length changes in early and late activated regions. As a consequence, the efficiency of cardiac pump function (the amount of stroke work generated by a unit of oxygen consumed) is approximately 30% lower in asynchronous than in synchronous hearts. Moreover, the amount of work performed in myocardial segments becomes considerably larger in late than in early activated regions. Cardiac Resynchronization Therapy (CRT) improves mechano-energetics of the previously asynchronous heart in various ways: it alleviates impediment of the abnormal contraction on blood flow, it increases myocardial efficiency, it recruits contraction in the previously early activated septum and it creates a more uniform distribution of myocardial blood flow. These factors act together to increase the range of cardiac work that can be delivered by the patients’ heart, an effect that can explain the increased exercise tolerance and quality of life reported in several CRT trials.

Three-dimensional simultaneous strain-volume analysis describes left ventricular remodelling and its progression: a pilot study

AIMS

Three-dimensional (3D)-echocardiography speckle imaging allows the evaluation of frame-by-frame strain and volume changes simultaneously. The aim of the present investigation was to describe the strain-volume combined assessment in different patterns of cardiac remodelling.

METHODS AND RESULTS

Fifty patients received a 3D acquisition. Patients were classified as follows: healthy subjects (CNT), previous AMI, and normal ejection fraction (EF; group A); ischaemic cardiomyopathy with reduced EF (group B); hypertrophic/infiltrative cardiomyopathy (group C). Values of 3D strain were plotted vs. volume for each frame to build a strain-volume curve for each case. Peak of radial, longitudinal, and circumferential systolic strain (Rεp, Lεp, and Cεp, respectively), slopes of the curves (RεSl, LεSl, CεSl), and strain to end-diastolic volume (EDV) ratio (Rε/V, Lε/V, Cε/V) were computed for the analysis. Strain-volume curves of the CNT group were steep and clustered, whereas, due to progressive dilatation and reduction of strains, progressive flattening could be demonstrated in groups A and B. Quantitative data supported visual assessment with progressive lower slopes (P< 0.05 for RεSl, CεSl, P= 0.06 for LεSl) and significantly lower ratios (P< 0.01 for Rε/V, Lε/V, and Cε/V). Group C showed an opposite behaviour with slopes and ratios close to those of normal subjects. Correlation coefficients between EDV and slopes of the curves were significant for all the directions of strain (CεSl: r = 0.891; RєSl: r = 0.704; LєSl: r = 0.833; P< 0.0001 for all).

CONCLUSION

We measured left ventricular volumes and strain by 3D-echo and obtained strain-volume curve to evaluate their behaviour in remodelling. A distinctive and progressive pattern consistent with pathophysiology was observed. The analysis here shown could represent a new non-invasive method to assess myocardial mechanics and its relationship with volumes.