Clinical Investigation
Echocardiography, Machine Learning, and Cardiac Resynchronization Therapy
Importance of Systematic Right Ventricular Assessment in Cardiac Resynchronization Therapy Candidates: A Machine Learning Approach

https://doi.org/10.1016/j.echo.2020.12.025Get rights and content

Highlights

  • Prediction of the impact of CRT on LV function and outcomes is often difficult.

  • CRT candidates are a highly heterogeneous population.

  • ML allows the analysis of a large amount of clinical and imaging data.

  • ML can identify clusters of patients with different characteristics and prognosis.

  • RV-derived features are important for the characterization of CRT candidates.

Background

Despite all having systolic heart failure and broad QRS intervals, patients screened for cardiac resynchronization therapy (CRT) are highly heterogeneous, and it remains extremely challenging to predict the impact of CRT devices on left ventricular function and outcomes. The aim of this study was to evaluate the relative impact of clinical, electrocardiographic, and echocardiographic data on the left ventricular remodeling and prognosis of CRT candidates by the application of machine learning approaches.

Methods

One hundred ninety-three patients with systolic heart failure receiving CRT according to current recommendations were prospectively included in this multicenter study. A combination of the Boruta algorithm and random forest methods was used to identify features predicting both CRT volumetric response and prognosis. Model performance was tested using the area under the receiver operating characteristic curve. The k-medoid method was also applied to identify clusters of phenotypically similar patients.

Results

From 28 clinical, electrocardiographic, and echocardiographic variables, 16 features were predictive of CRT response, and 11 features were predictive of prognosis. Among the predictors of CRT response, eight variables (50%) pertained to right ventricular size or function. Tricuspid annular plane systolic excursion was the main feature associated with prognosis. The selected features were associated with particularly good prediction of both CRT response (area under the curve, 0.81; 95% CI, 0.74–0.87) and outcomes (area under the curve, 0.84; 95% CI, 0.75–0.93). An unsupervised machine learning approach allowed the identification of two phenogroups of patients who differed significantly in clinical variables and parameters of biventricular size and right ventricular function. The two phenogroups had significantly different prognosis (hazard ratio, 4.70; 95% CI, 2.1–10.0; P < .0001; log-rank P < .0001).

Conclusions

Machine learning can reliably identify clinical and echocardiographic features associated with CRT response and prognosis. The evaluation of both right ventricular size and functional parameters has pivotal importance for the risk stratification of CRT candidates and should be systematically performed in patients undergoing CRT.

Section snippets

Population

Two hundred nine patients with systolic HF undergoing CRT device implantation according to current guidelines1 at Oslo University Hospital (Norway), Leuven University Hospital (Belgium), Rennes University Hospital (France), Aalst OLV Hospital (Belgium), and Karolinska University Hospital (Sweden) between August 2015 and November 2017 were prospectively included in this observational, multicenter study. Sixteen patients were excluded from the final analysis because of study withdrawal (n = 4),

Results

The baseline characteristics of the population are depicted in Table 1.

The population had a mean age of 67 years. The majority of patients were men (70%), with a high prevalence (67%) of nonischemic dilated cardiomyopathy. One hundred sixty-eight patients (87%) had typical LBBB.

Discussion

In this prospective, multicentric study, we have shown the feasibility and validity of the application of both supervised and unsupervised ML approaches to a comprehensive pattern of preimplantation clinical, biochemical, electrocardiographic, and echocardiographic data obtained in everyday clinical practice in CRT candidates. Our analysis allowed the identification of groups of features that are significantly associated with CRT response and prognosis and the identification of two phenogroups

Conclusion

The application of ML methods to a set of common clinical, electrocardiographic, and echocardiographic data emphasizes the importance of a multiparametric approach for both the identification of CRT response and the prediction of prognosis after CRT. Our results underscore the importance of RV function on both CRT response and outcome and the pivotal role of the global assessment of heart function in patients undergoing CRT. Despite these interesting results, additional studies on a broader

References (36)

  • M. Brignole et al.

    2013 ESC guidelines on cardiac pacing and cardiac resynchronization therapy: the Task Force on Cardiac Pacing and Resynchronization Therapy of the European Society of Cardiology (ESC). Developed in collaboration with the European Heart Rhythm Association (EHRA)

    Europace

    (2013)
  • E.S. Chung et al.

    Results of the Predictors of Response to CRT (PROSPECT) trial

    Circulation

    (2008)
  • S. Yang et al.

    A novel risk model for mortality and hospitalization following cardiac resynchronization therapy in patients with non-ischemic cardiomyopathy: the alpha-score

    BMC Cardiovasc Disord

    (2020)
  • D. Patel et al.

    Predictors of long-term outcomes greater than 10 years after cardiac resynchronization therapy implantation

    J Cardiovasc Electrophysiol

    (2020)
  • W. Zareba et al.

    Effectiveness of cardiac resynchronization therapy by QRS morphology in the Multicenter Automatic Defibrillator Implantation Trial-Cardiac Resynchronization Therapy (MADIT-CRT)

    Circulation

    (2011)
  • J. Lumens et al.

    Differentiating electromechanical from non-electrical substrates of mechanical discoordination to identify responders to cardiac resynchronization therapy

    Circ Cardiovasc Imaging

    (2015)
  • S.H. Shah et al.

    Opportunities for the cardiovascular community in the precision medicine initiative

    Circulation

    (2016)
  • M. Cikes et al.

    Machine learning-based phenogrouping in heart failure to identify responders to cardiac resynchronization therapy

    Eur J Heart Fail

    (2019)
  • Cited by (12)

    • Contemporary Applications of Machine Learning for Device Therapy in Heart Failure

      2022, JACC: Heart Failure
      Citation Excerpt :

      In addition to biventricular dyssynchrony, the role of right ventricular (RV) dysfunction in predicting outcomes after CRT is still debatable.33,34 A recent ML approach by Galli et al28 combined clinical, electrocardiographic, LV, and RV echocardiography parameters to predict CRT response and outcomes. Out of the 16 variables that were predictive of CRT response, 8 pertained to RV dynamics.

    • Future Guidelines for Artificial Intelligence in Echocardiography

      2022, Journal of the American Society of Echocardiography
    • Prediction of response after cardiac resynchronization therapy with machine learning

      2021, International Journal of Cardiology
      Citation Excerpt :

      Intriguingly, with multiple distinct algorithms we reached similar results showing that linear models performed better in prediction than nonlinear ones. Moreover, two recent studies reported construction of ML predictive models of CRT response incorporating more echocardiographic features, some particularly correlated to LV mechanical dyssynchrony, and achieved AUCs of over 0.8 [27,28]. These results highlighted the promising future of integrating the strength of ML techniques and features specifically correlated to LV dyssynchrony to further improve predictive performance.

    • Artificial Intelligence and Machine Learning in Electrophysiology—a Short Review

      2023, Current Treatment Options in Cardiovascular Medicine
    View all citing articles on Scopus

    Drs. Galle and Le Rolle contributed equally to this work.

    Dr. Aalen was supported by a grant from the Norwegian Health Association. Dr. Larsen was a recipient of a clinical research fellowship from the South-Eastern Norway Regional Health Authority. Dr. Voigt holds a research mandate from the Research Foundation Flanders (FKM1832917N). Dr. Le Rolle was supported by the French National Research Agency (ANR-16-CE19-0008-01, project MAESTRo).

    Conflicts of interest: None.

    View full text