A team led by Professor Nicholas Mills is developing AI decision-support tools to diagnose acute heart conditions such as heart attack in hospital Emergency Departments.
Every day, people arrive at hospital Emergency Departments suspected of having a heart attack. But the symptoms are nonspecific: chest pains and shortness of breath have a variety of causes. Diagnostic tests are needed to identify those who are having a heart attack and should stay in hospital for further treatment and those who can safely go home.
Heart attacks are usually diagnosed with serial blood tests that measure the levels of cardiac troponin, which is released by the heart muscle in response to injury and inflammation. These blood tests are supposed to be carried out at specific time points, which is often infeasible in busy hospitals, and they use the same thresholds for every patient. However, various factors affect blood troponin levels, including the time since the onset of symptoms, the age and sex of the patient, and the presence of other health problems. This means blood troponin tests are not always accurate, which can lead to misdiagnosis and inequalities in care, ultimately delaying treatment, as well as hospital admission and overtreatment of people who are not having a heart attack.
To individualise diagnosis and improve accuracy, a team led by Professor Nicholas Mills, British Heart Foundation Chair of Cardiology at the University of Edinburgh and a Consultant Cardiologist at the Royal Infirmary of Edinburgh, took a data-driven and advanced computational approach.
With funding from the National Institute for Health and Care Research (NIHR) and the British Heart Foundation, alongside support from Edinburgh Innovations, they developed and patented an AI decision-support tool: CoDE-ACS (Collaboration for the Diagnosis and Evaluation of Acute Coronary Syndrome). This uses anonymised data from digitised healthcare records, including a patient's age, sex, their electrocardiogram (ECG) results and medical history, as well as blood troponin levels (which do not need to be serially measured). Initially developed using this information from 10,038 patients in Scotland who had arrived at hospital with a suspected heart attack, the tool predicts the probability that an individual is having a heart attack, giving a score from 0 to 100 to aid clinical judgment.
Testing the performance of CoDE-ACS on a further 10,286 patients revealed excellent discrimination for heart attack, with an accuracy of 99.6%, outperforming current care pathways that use fixed thresholds of blood troponin levels. Moreover, the AI tool performed well regardless of age, sex, or pre-existing health conditions, showing its potential for reducing misdiagnosis inequalities.
The team is receiving support from Edinburgh Innovations, whose business development, technology transfer and venture creation teams are helping Professor Mills advance this invention towards clinical application by providing tailored advice, identifying translational opportunities, brokering partnerships and supporting the development of a commercialisation strategy. Together, they are working to develop CoDE-ACS into a standalone regulatory-compliant Software as Medical Device that can be commercialised and deployed into the clinic.
The innovative approach of Professor Mills and his team could be applied to a variety of conditions, harnessing AI to analyse multiple types of data and develop tools that stratify patients most at risk and likely to benefit from certain treatment pathways. Indeed, they have used the same principles to develop a machine learning tool that assesses age and blood levels of a heart hormone that is released in response to stress or excessive pressure, identifying people suffering acute heart failure (Collaboration for the Diagnosis and Evaluation of Heart Failure, CoDE-HF). As Professor Mills says,
Harnessing data and artificial intelligence to support clinical decisions has enormous potential to improve care for patients and efficiency in our busy emergency departments.”
