Funktionelle und prognostische Bedeutung Cardiovascular Magnetic Resonance Feature Tracking basierter Analyse des myokardialen Remote Strains nach akutem Myokardinfarkt

Functional and prognostic significance of Cardiovascular Magnetic Resonance Feature Tracking based analysis of the myocardial remote strain after acute myocardial infarction

by Patricia Charlotte Boom
Doctoral thesis
Date of Examination:2022-09-13
Date of issue:2022-12-20
Advisor:Prof. Dr. Dr. Andreas Schuster
Referee:Prof. Dr. Dr. Andreas Schuster
Referee:PD Dr. Johannes Kowallick
Sponsor:Deutsches Zentrum für Herz-Kreislauf-Forschung e.V.
Persistent Address: http://dx.doi.org/10.53846/goediss-9468

 

 

Files in this item

Name:Dissertation Patricia Boom.pdf
Size:1.56Mb
Format:PDF

The following license files are associated with this item:

Abstract

English

Background: Cardiovascular disease is the leading cause of death worldwide. Nowadays, the risk stratification after an acute myocardial infarction (AMI) is mostly realised by echocardiography. However, the importance of cardiac magnetic resonance imaging (CMR) continues to increase as it allows a better assessment of the function and condition of the myocardium. Former studies have shown that CMR feature tracking (CMR-FT)-derived global strain assessments provide incremental prognostic information in patients following AMI. One aim of this study was to determine if the regional strain within the remote non-infarcted myocardium (RNM) constitutes a significant risk factor that adds prognostic value after AMI. Furthermore, the study focused on the comparison of a conventional and an artificial intelligence (AI)-based postprocessing software in order to investigate feasibility and prognostic implications of AI-based scar quantification. Methods: In total 1,168 AMI patients from two myocardial infarction multicenter trials were included. Regarding the outcome of the patients, major adverse clinical events (MACE) were recorded during a follow-up of 12 months comprising death, reinfarction and congestive heart failure. CMR-FT-derived regional strain analyses as well as infarct size evaluation and identification of RNM were performed within a cohort of 1,034 patients. Moreover, a total of 913 patients were manually and automatically assessed using conventional and AI-based software for scar quantification. Results: Patients with MACE had significantly lower RNM circumferential strain (CS) than those without MACE. Additionally, impaired RNM CS was a strong predictor of MACE (HR 1.05, 95% CI 1.07-1.17, p = 0.003). A cut-off value for RNM CS of - 25.8% best identified high-risk patients (p < 0.001). Especially among patients who are already considered to be at risk for MACE due to established parameters, such as a reduced left ventricular ejection fraction of less than 35%, the analysis of RNM CS enabled a profound dichotomization into high and low risk patients (p = 0.038). Regarding the AI-based scar quantification, a manual plotting of the microvascular obstruction (MVO) preceded the analysis. Comparison of automated infarct size after plotting the MVO and manually derived infarct size revealed considerable differences in absolute numbers (p < 0.001). After manual correction of the scar and myocardial borders, no significant differences between the automatic and manual infarct sizes were determined (p = 0.07). In the ROC analysis concerning the prediction of MACE no significant differences between the automatic approach after manual plotting of the MVO compared to the conventional software were observed (AUC automatic approach 0.65, AUC manual approach 0.66, p = 0.72). Conclusion: RNM CS is a useful parameter to characterize the response and function of the remote myocardium and underpins the compensatory performance of the viable myocardium. Moreover, it allows improved stratification following AMI especially in patients, who already belong to a high-risk group for MACE due to acknowledged clinical parameters. The AI-based software for infarct size quantification seems promising for optimized time-efficient analyses but cannot fully be established in everyday clinical practice yet, before implementing automated MVO detection within the scar.
Keywords: Cardiac magnetic resonance imaging; Feature Tracking; Artificial intelligence; acute myocardial infarction; Myocardial scar
Schlagwörter: Kardiale Magnetresonanztomografie; Akuter Myokardinfarkt; Myokardiale Narbe
 
Statistik