Myocardial Ischemia

Stylised image of a human heart
Stylised image of a human heart

Academic Lead: Dr Amedeo Chiribiri

Clinical Area: Cardiovascular

Partner: Siemens Healthineers, Philips Healthcare, Circle cardiovascular Imaging, Medis, GE Healthcare

Heart disease is the second most common cause of death and reason for hospital admissions in the NHS for patients aged 75 or older. Coronary artery disease (CAD) was listed as the most frequent cause of these acute heart disease admissions to hospital. CAD develops when the major blood vessels that supply the heart with blood, oxygen and nutrients become damaged. This causes a decreased blood flow to the heart which can cause chest pain and shortness of breath. If the blood vessels are completely blocked this can cause a heart attack.  

To assess blood flow through the heart, a small tube can be inserted into the blood vessels of the heart where the dye is injected and an X-ray image is taken which outlines any areas in the vessel that are narrow or blocked. This is considered an invasive strategy. The doctor will decide based on these images if the patient needs treatment, which can be performed using a balloon inserted through the same tube and inflated to improve the blood flow in the vessels. A meshed tube (stent) may then be used to keep the dilated vessel open. 

Sometimes the doctor may ask for a stress test which can involve taking images of the heart while on a treadmill ride or after medications are given to simulate the heart being used more than when it is relaxed (e.g. when sitting normally). This is considered a non-invasive strategy. The imaging methods used for these non-invasive tests available include ECG (heart tracing), echocardiogram (ultrasound scan of the heart), nuclear perfusion scan or magnetic resonance scan (MRI). MRI has become the gold standard imaging method for these stress tests due to its accuracy and consistent results.  Cardiac MRI uses the same type of radio waves as mobile phones do to take snapshots of the heart of the patients, without exposure to harmful ionising radiations. 

The interpretation of the results of these tests sometimes relies on the availability of trained experts. However, these specialists usually only work in tertiary hospitals, which means there may not be access to this testing in smaller or rural parts of the UK. In addition, sometimes different tests produce different results and the physician is faced with a decision on which test to trust. There are also some guidelines that are open to interpretation and can depend on the clinician available. These factors result in suboptimal outcomes and increased societal and economic costs. 


We will use deep-learning artificial intelligence on patients with known or suspected myocardial ischaemia to improve the personalisation of healthcare. vg5New advances in medicine have the potential to revolutionise our approach to CAD particularly by applying with Artificial Intelligence (AI) techniques. This project focuses specifically on the use of AI to unravel the complex relationship between medical images and patients’ symptoms. Our ultimate goal is to provide personalised care to our patients which does not rely on the local availability of resources and expertise but can be provided by AI-enabled software deployed locally or on the NHS cloud. Our patients will benefit by receiving the best diagnosis, leading to the best choice of treatment, by reducing complication rates, particularly in case of non-appropriate interventions. 

In this project, we will use our rich large scale historical dataset at St Thomas’ Hospital - which contains x-ray coronary angiograms, stress perfusion MRI scans, nuclear medicine scans, non-invasive coronary CT scans (CTCA), stress echo images – to develop novel algorithms that can guide physicians in selecting right from the start the best treatment pathway. The project will benefit from the capability of AI algorithms developed on large populations of subjects in detecting features or patterns that are beyond human eye appreciation, to build prediction models able to given each coronary lesions its significance in terms of how much it is affecting the heart muscle, how much it is linked to patients’ symptoms and to life expectancy. We believe that the end product of the study will be of great interest to any of the potential industrial partners listed above.