Long-term outcomes of post-infarction aneurysm repair and preoperative prediction for left ventricular resection: A non-randomized retrospective cohort study

Background. Reverse left ventricular remodeling is a process characterized by a reduction in cardiac chamber volume and the restoration of normal ventricular geometry. These structural changes enhance myocardial systolic and diastolic function, specifically improving the heart’s ability to efficiently fill and eject blood. However, some patients may still experience the clinical progression of circulatory insufficiency. Postoperative reverse cardiac remodeling is influenced by optimal therapy, coronary revascularization, left ventricular aneurysm repair, mitral regurgitation correction, and cardiac resynchronization therapy. Objective. To evaluate surgical outcomes in patients with post-infarction left ventricular aneurysms and to develop an artificial intelligence model for predicting postoperative echocardiographic parameters across different follow-up periods. Methods. This non-randomized retrospective cohort study evaluated 130 patients with post-infarction left ventricular aneurysms who had undergone the Dor procedure and coronary artery bypass grafting. The surgeries were performed at the Research Institute – Specialized Cardiac Surgery Clinical Hospital Named after Academician B.A. Korolev in 2011–2021. The primary study endpoints included the predictors of complications associated with hemodynamic correction during both the early and long-term postoperative follow-up, as well as the feasibility of using an artificial intelligence model (CatBoost) to determine the extent of left ventricular myocardial resection. The model was trained using retrospective data on short- and long-term surgical outcomes. The following preoperative echocardiographic parameters served as predictors: end-diastolic volume, end-systolic volume, ejection fraction, and sphericity index. The target variables consisted of the same parameters recorded 7 days postoperatively. During training, optimal hyperparameter values were selected for each target variable to ensure the best predictive performance for Day 7. Similar models were constructed to predict outcomes at 3 and 6 months after surgery. Results. Preoperative echocardiography demonstrated that cardiac volumetric parameters were significantly elevated: end-diastolic volume was 169 [141–205] mL; end-systolic volume, 95[75–128] mL; ejection fraction, reduced to 41.7%; and the sphericity index, elevated at 0.7 [0.7–1]. These data indicate initial signs of left ventricular remodeling. The study showed a statistically significant reduction in left ventricular volumetric parameters in patients during both the early and long-term postoperative follow-up. The artificial intelligence model (CatBoost) yielded a low median prediction error (20%) for the estimated cardiac parameters compared to actual values. This demonstrates high accuracy of the model, as well as its potential clinical utility for surgical planning and complication risk reduction. Conclusion. During long-term postoperative follow-up, patients exhibited significant reverse left ventricular remodeling, as evidenced by positive changes in volumetric parameters and ejection fraction. This predictive model constitutes a promising tool for preoperative planning, specifically for optimizing surgical techniques and reducing the risk of low cardiac output syndrome.