Deep Learning Applications in Ultrasound Imaging for Cardiovascular Assessment

Deep learning has revolutionised ultrasound imaging for cardiovascular assessment by automating the extraction of complex features and enhancing diagnostic precision. Traditional ultrasound examinations often rely on operator expertise for tasks such as boundary delineation and plaque detection, leading to variability and time constraints. Convolutional neural networks have been employed to segment vessel walls, quantify intima-media thickness and identify atherosclerotic plaques with high accuracy. Architectures such as U-Net and attention-based networks can process two- and three-dimensional ultrasound data, yielding reproducible measurements while reducing user interaction. Transfer learning further refines performance in settings with limited labelled data, and emerging explainable AI techniques improve clinical trust by visualising model decisions. These advances enable real-time, operator-independent analysis for early detection of vascular disease, longitudinal monitoring of plaque progression and comprehensive evaluation of cardiac function. Integration into portable and cloud-based platforms promises to expand access to advanced cardiovascular diagnostics across diverse healthcare environments worldwide.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Deep Learning Applications in Ultrasound Imaging for Cardiovascular Assessment publication trend

The graph below shows the total number of articles in deep learning applications in ultrasound imaging for cardiovascular assessment across all publications each year (not limited to Nature Index journals).

Technical terms

Deep learning: Machine learning technique using multi-layered neural networks to automatically learn hierarchical representations from data.

Convolutional neural network (CNN): A specialised deep learning architecture employing convolutional layers to detect spatial patterns and features in images.

Transfer learning: Method of adapting a pre-trained model to a new but related task, reducing the need for large annotated datasets.

Segmentation: Process of partitioning an image into meaningful regions, such as vessel walls or plaques, for quantitative analysis.

Dice similarity coefficient (DSC): Statistical metric measuring the overlap between two segmented regions, indicating accuracy of segmentation.

Three-dimensional ultrasound (3DUS): Imaging modality that acquires volumetric ultrasound data, enabling detailed spatial assessment of cardiovascular structures.

References

1. Deep learning‐based carotid media‐adventitia and lumen‐intima boundary segmentation from three‐dimensional ultrasound images. Medical Physics (2019).
2. Unseen Artificial Intelligence—Deep Learning Paradigm for Segmentation of Low Atherosclerotic Plaque in Carotid Ultrasound: A Multicenter Cardiovascular Study. Diagnostics (2021).
3. Attention-Based UNet Deep Learning Model for Plaque Segmentation in Carotid Ultrasound for Stroke Risk Stratification: An Artificial Intelligence Paradigm. Journal of Cardiovascular Development and Disease (2022).