Artificial Intelligence-Based Solutions for Supporting Cardiovascular Disease Diagnosis

Cardiovascular disease remains the leading cause of mortality worldwide, placing a significant burden on healthcare systems. Despite substantial advancements in medical diagnostics and therapeutic interventions, the clinical assessment of heart failure still relies on traditional imaging techniques, such as echocardiography. Among various diagnostic parameters, Left Ventricular Ejection Fraction (LVEF) is a crucial metric for evaluating systolic cardiac function and stratifying heart failure phenotypes. However, its manual estimation depends on operator expertise, leading to inter- and intra-observer variability that compromises diagnostic reproducibility. To address these limitations, the developed approach aims to improve the precision and reliability of LVEF quantification while reducing the dependency on manual interpretation. This study proposes an automated framework for estimating LVEF from echocardiographic images using AI-driven methodologies. Specifically, the research focuses on implementing the Area-Length method to compute left ventricular end-diastolic volume (EDV) and end-systolic volume (ESV) from apical four-chamber echocardiographic views. Additionally, a comparative statistical analysis is conducted to evaluate the variability of the automated method against ground truth annotations from the EchoNet-Dynamic dataset, a publicly available repository of labelled echocardiographic videos. This research contributes to the development of CardioSmartAssist, a clinical decision-support software designed to streamline the assessment of LVEF for cardiologists. By enhancing the reproducibility of echocardiographic analysis, this work facilitates more efficient heart failure diagnostics. The findings highlight the potential of AI-driven solutions to optimize clinical workflows, offering a promising avenue for the future of automated cardiovascular imaging.