Active Computing & AI Heart, Stroke & Blood

UNITY: UK collaborative for integrating AI into echocardiography

In plain English

AI plain-English summary

Echocardiography—ultrasound scans of the heart—produces images that are notoriously difficult for artificial intelligence to interpret, because the heart’s borders are blurry and measurements must be precise. A UK-wide team of cardiologists, physiologists, and AI experts aims to change that. They have already built and published AI models trained on tens of thousands of expert-labelled echo images. Now they will tackle seven specific hurdles: creating a national bank of expert annotations; designing AI tailored to echo’s unique visual challenges; automating the identification of cardiac cycle timepoints; making Doppler measurements; standardising clinical acquisition protocols; validating AI against UK experts; and comparing results with cardiac MRI. If successful, this programme could give the NHS a validated, open AI tool that automates the most time-consuming parts of echo analysis—freeing clinicians to focus on diagnosis and treatment rather than manual measurement. It would also provide the UK’s many world-leading echo researchers with a shared infrastructure for future studies. The work is directly translational, aligned with the NHS’s five-year plan to become a world leader in clinical AI.

View original technical description
Echocardiography is the bedrock of assessing the heart in clinical practice. There is huge potential for Artificial Intelligence (AI) to automate analysis, increase efficiency, and tackle image quality and reproducibility issues. However, echo is more challenging than conventional AI image processing (e.g. for self-driving cars) because borders are less distinct, yet precision more important, and specific measurements are the focus. Our UK Collaborative of cardiologists, physiologists and AI experts, and the British Society of Echocardiography, will take AI in echo forward into clinical applicability, aligned with the NHS plan for world leadership in AI in 5 years. Our pilot work has already delivered tens of thousands of expert labels, from which we have successfully built, presented and published AIs. Our programme addresses 7 needs for AI in echo: (1) a national bank of annotations by named experts on tens of thousands of echo images; specially designed AI technology for (2) echo images, (3) identifying cardiac cycle timepoints, (4) making Doppler measurements; (5) clinical acquisition protocols to improve reproducibility; (6) open validation against UK experts; (7) comparison with cardiac MR. It will also provide the UK’s many world authorities in echo with tools for their own future research.

View the original record at the funder ↗

Researchers

Darrel Francis (EPMC Awardee)

Related Research

Grants with similar aims, by meaning.

AI-enabled ECG to reduce demand for echocardiography and shorten diagnostic waiting lists
Interactive AI-mentored training system to deliver quality-assured echocardiographers (Ms Catherine Stowell)
BHF Rosetrees Fellowship: Using AI for qualitative echocardiography judgements
Artificial intelligence to improve the diagnosis of acute cardiovascular conditions
Prospective Longitudinal evaluation of AI-ECG in a Newly diagnosed Heart Failure (PLANE-HF)

Original classification

Programme Grant

Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.