Showcasing Artificial Intelligence for #HeartMonth

To recognise and celebrate British Heart Foundation’s #HeartMonth, the AI Centre is showcasing our current research projects involved in streamlining heart disease diagnosis and treatment. ?

Heart disease affects millions of people nationwide and puts immense pressure on NHS resources and staff. Clinical artificial intelligence tools can help streamline parts of the diagnostic-treatment pathway, thereby freeing up valuable resources such as hospital staff time for other important tasks and face-to-face patient time.

Alistair Young, Professor of Cardiovascular Data Analytics and AI at King’s College London, School of Biomedical Engineering & Imaging Sciences, is technical lead on one of the AI Centre funded projects on cardiovascular clinical AI.

“The role of AI in heart disease is not to replace the doctor, but to assist them in their diagnostic and prognostic decisions. Doctors are under enough time and resource pressure as it is, but we can alleviate some of this burden with clinical AI tools that streamline processes.”

Alistair Young, Professor of Cardiovascular Data Analytics and AI, School of Biomedical Engineering & Imaging Sciences, King’s College London?

Professor Young outlines the original clinical problem identified, the aim of his team’s research, and the anticipated applications of the clinical AI tool he is working on. This research focuses on atrial fibrillation – one of the most common forms of cardiac arrhythmia (otherwise known as irregular heartbeat). It affects more than 33 million people worldwide - and increases the risk of stroke, heart failure, and death.

What was the clinical issue, or unmet need, driving this research project?

Identifying the geometry and volume of the left atrium (heart chamber) is incredibly useful in confirming diagnosis and prognosis of various heart diseases, including atrial fibrillation. Currently, the left atrial volume is estimated using 2 different measurements - width and length. This method is somewhat limited and can result in inaccurate calculations, and therefore poor clinical decisions.

The geometry and volume of the left atrium can inform the treatment procedure for atrial fibrillation. The most common treatment for atrial fibrillation is catheter ablation – it involves inserting a flexible catheter into the blood vessels and using local heating or freezing to destroy the abnormal tissue causing the atrial fibrillation. Unfortunately, this procedure has a high re-occurrence rate of 20-50%. This is partly due to the limited and inaccurate information clinicians have on the left atrial geometry and volume.

How does your research address this?

We are developing an AI tool, in collaboration with Siemens Healthineers, for the automatic analysis of the heart chambers from MRI images. The AI tool will infer the 3D shape, volume, and surface area of the left atrium, to a greater degree of accuracy than the current bi-plane method that uses only 2 measurements.

The AI tool will be able to provide clinicians with the 3D results, reducing the staff resources need to calculate these left atrial measurements, and identifying patients who would benefit from further investigation.

How will this AI tool be used?

The AI tool would be implemented in the current clinical pathway, so that the resulting information report is available almost instantaneously for the clinician to use at the time of imaging. In the future, the results will be extended to include risk prediction scores and highlight sites appropriate for ablation treatment.

For more context on this area of research, you can access the related academic paper entitled “Deep Learning Estimation of Three-Dimensional Left Atrial Shape from Two-Chamber and Four-Chamber Cardiac Long Axis Views” here.

Amedeo Chiribiri, Professor in Cardiovascular Imaging and Consultant Cardiologist, and Avan Suinesiaputra, Research Associate of Cardiovascular Imaging are currently working on the second of our two-heart disease research project. Their project focuses primarily on coronary artery disease (CAD). CAD is hugely complex, and the landscape of diagnostics, clinical guidelines, diagnostic and therapeutic steps, and therapy and treatment options are always rapidly changing. ?

“Currently, MRI and CT imaging are used to diagnose coronary artery disease that can present as chest pain. While these non-invasive imaging techniques can identify and clear most patients of significant disease, there are hundreds of images produced that require extensive analysis and radiologist hours. This ends up being a major bottleneck in clinical decision-making.”

Amedeo Chiribiri, Professor of Cardiovascular Imaging and Consultant Cardiologist, School of Biomedical Engineering & Imaging Sciences, King’s College London?

Amedeo and Avan’s research project aims to use AI tools to improve the process of streamlining referrals in patients with suspected CAD. The AI tool will select the most appropriate scanning techniques; thereby reducing scan duration and improving patient experience. Whilst this AI tool is not yet being used clinically, examples of applications include decision making around:

1. Stressing the patient – to avoid unnecessarily uncomfortable procedures.
2. Acquiring full resolution images – minimising unnecessary use of hospital resources.
3. Repeating image scanning techniques – minimising unnecessary use of both hospital resources and staff time.

The research projects we have chosen to showcase for #HeartMonth primarily focus on improving efficiency and accuracy, but AI also has the potential to:?

• Improve early detection: One of the key challenges with heart disease is that it often has no noticeable symptoms until it is in an advanced stage, making early treatment a challenge. AI tools can analyse vast amounts of hospital and imaging data to identify patterns and risk factors that may point towards heart disease. Using these tools, clinicians will be able to intervene early and prevent the progression of the disease.
• Personalise treatment plans: AI tools can personalise treatment plans by looking at medical history, genetic data, and other relevant information and promptly designing a plan that would otherwise require a team of specialists to meet and discuss, freeing up Consultant time.

If you are interested in exploring other disease area research projects, visit the AI Centre website.