AI Cardia : Smart Detection of Heart Conditions

Abstract

Cardiovascular diseases (CVDs) remain the leading cause of death worldwide, empha sizing the need for timely and accurate diagnosis. Cardiac MRI is the gold standard for assessing heart conditions, but the manual interpretation of these images is a complex, time consuming process requiring specialized expertise. Cardiologists and radiologists must manually analyze MRI scans to evaluate heart function, segment key regions, and compute critical cardiac parameters such as End-Systolic Volume (ESV), End-Diastolic Volume (EDV), and Ejection Fraction (EF). This manual approach introduces the risk of delayed diagnoses, human error, and variability in results. In critical cases, such delays could mean the difference between life and death, as immediate intervention is often required for patients at risk of heart failure or cardiac arrest. To address these challenges, this project aims to develop an AI-powered mobile application that automates cardiac MRI analysis, providing diagnostic support to healthcare profes sionals. The system will segment the left ventricle and myocardium, compute essential cardiac metrics, and classify heart conditions to assist in clinical decision making. The project will choose and utilize datasets from well established cardiac MRI repos itories. The MICCAI 2009 Sunnybrook dataset consists of MRI DICOM images from 45 patient batches, accompanied by coordinate files and patient data for precise segmentation and measurement. Alternatively, the ACDC dataset contains 1,800–2,000 MRI data files, enabling the classification of heart conditions such as normal, dilated cardiomyopathy, hypertrophic cardiomyopathy, and myocardial infarction. This project is aimed at strengthening the accuracy, efficiency, and accessibility of cardiac MRI analysis by incorporating AI-based automation in the diagnosis process so that crit ical diagnoses could be performed in time and therefore, improve patient outcomes and decrease the workload of healthcare professionals