Presented by Ikteder Akhand Udoy
Computing PhD, Data Science emphasis
Hybrid Location: MEC 201 and via Zoom
Abstract: Cardiovascular diseases (CVD) are a leading cause of mortality worldwide, and timely, accurate diagnosis remains critical to improving patient outcomes. Traditional diagnostic tools such as electrocardiograms (ECGs) and echocardiography rely heavily on clinician expertise, often leading to variability and potential delays. This presentation explores recent advancements in artificial intelligence (AI) and deep learning that have the potential to transform cardiovascular diagnostics. Synthesizing findings from key studies, we focus on the application of deep learning models—including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and hybrid architectures—across medical imaging and time-series data for enhanced CVD detection. Notable breakthroughs include EchoNet-Dynamic, a real-time, video-based AI model achieving accuracy levels that surpass human experts in cardiac function assessment, and hybrid CNN-RNN models that improve ECG interpretation. This presentation will also address data processing schemes, model architectures, and results from computational artifacts that assess model performance across datasets like PneumoniaMNIST and BreastMNIST. Finally, we will discuss current challenges, including data generalizability and clinical integration, and propose future research directions toward building scalable, AI-driven diagnostic solutions that support efficient and accurate cardiovascular healthcare.
Committee: Dr. Omiya Hassan (chair), Dr. Nasir Eisty, Dr. Kurtis Cantley, Dr. Oliviero Andreussi (Comp EE)