Dissertation

Automated Solutions for Cardiovascular Magnetic Resonance Imaging Analysis

Hua, Chong Jun

2022

Overview

Cardiovascular magnetic resonance (CMR) is an important cardiac imaging tool for assessing the prognostic extent of myocardial injury after myocardial infarction (MI) and for assessing the efficacy of potential cardioprotective therapies in reducing MI size and preventing adverse left ventricular (LV) remodelling in reperfused MI. However, manual contouring and analysis can be time-consuming with interobserver and intraobserver variability, which can in turn lead to reductions in accuracy and precision of analysis. There is thus a need to automate CMR scan analysis in MI patients to save time, increase accuracy, reproducibility and precision. Automated imaging analysis techniques based on artificial intelligence (AI) that are developed with machine learning (ML), and more specifically deep learning (DL) strategies, can enable efficient, robust, accurate and clinician-friendly tools to be developed to improve both clinician productivity and quality of patient care. In this regard, we have identified the analysis of MI CMR scans as a use case to develop our prototype solution for automated CMR analysis. In order to train, test and validate our automated model, we used MI CMR scans from the CONDI-2/ERIC-PPCI and IMMACULATE trials. Remote ischaemic conditioning (RIC) with transient ischaemia and reperfusion applied to the arm has been shown to reduce MI size in patients with ST-elevation myocardial infarction (STEMI) undergoing primary percutaneous coronary intervention (PPCI), although the CONDI-2/ERIC-PPCI trial failed to report any benefit of RIC on the incidences of cardiac death and hospitalisation for heart failure at 12 months following STEMI. We discuss the scientific basis of RIC and describe the cardioprotective potential of RIC and chronic RIC (CRIC) in various cardiac conditions including STEMI and cancer therapeutics-related cardiotoxicity. The effect of limb RIC on MI size and left ventricular ejection fraction (LVEF) was investigated in a pre-planned CMR substudy of the CONDI-2/ERIC-PPCI trial. We summarise the results of this trial and describe the journey towards attainment of manual analysis gold standard through expert observer analysis of MI CMRs from this substudy. Finally, manually annotated MI CMRs from the CONDI-2/ERIC-PPCI substudy were used as testing and training data for the development of an automated analysis solution prototype, with a separate cohort of studies from the IMMACULATE study being used for internal validation. This has also been summarised in the graphical abstract below.In summary, we have shown that AI, ML and more specifically DL can improve efficiency and accuracy in the assessment of important prognostic imaging biomarkers in MI CMR scans through the development of an automated CMR analysis tool called CardiacApp. Global collaboration through open-source publication of codes and datasets will help to further advance the field and improve governance, safety and quality control through knowledge sharing. Future directions include the incorporation of automated analysis within clinical trials that assess clinical outcomes, and the extension of existing ML applications to a wider range of imaging biomarker analysis within heterogenous patient cohorts.

• Publisher: ProQuest Dissertations & Theses
• Subject: Medical imaging
• ISBN: 9798291508336