Review paper on deep learning architectures for cardiac image analysis published in JACC: Cardiovascular Imaging

A new review paper, led by Joske van der Zande, has been published in JACC: Cardiovascular Imaging. The paper provides a comprehensive overview of advanced deep learning (DL) architectures applied to cardiac image analysis. The authors systematically reviewed recent research on state-of-the-art DL approaches, including transformers, foundation models, and neural network compression techniques. It covers a wide range of cardiac imaging tasks such as segmentation, classification, and disease detection across multiple modalities. It also surveys the most commonly used publicly available cardiac imaging datasets and analyzes recent contributions focused on deep model compression — an increasingly important topic for deploying AI in clinical environments with limited computational resources.

Key findings highlight that DL techniques have seen significant progress in cardiac image analysis, particularly through the adoption of attention-based architectures and foundation models. The paper concludes with a critical discussion of open challenges in the field and outlines promising directions for future research, including robustness, generalizability, and integration into clinical workflows.

👉 Read the full review here!