The role of artificial intelligence in cardiology
From heart monitoring and preventing heart attacks to assessing heart damage, AI has tremendous potential in reducing cardiovascular mortality
ARTIFICIAL Intelligence (AI) is transforming various fields, with healthcare being one of the most impacted. AI involves the development of computer systems capable of learning, reasoning and emulating human intellectual abilities. As AI advances rapidly, some experts anticipate that human-AI singularity could occur within the next 20 to 30 years.
Regardless of this outcome, AI is poised to play a pivotal role in healthcare, with cardiology at the forefront of this transformation. Significant progress has been made over the past decade in applying AI to heart disease.
Home heart monitoring
Envision a future where an affordable wearable device allows you to monitor your heart health at home. This device could transmit biophysical signals to an AI system, which would assess and predict your risk of future heart events. AI advances in electrocardiography (ECG) are bringing this vision closer to reality.
ECG is a widely used, cost-effective, non-invasive diagnostic tool that records the heart’s electrical signals. Although automated ECG interpretation is common in medical centres, it often requires oversight from specialists. Recent advancements in AI have led to ECG applications that exceed the performance of American-certified heart rhythm specialists in detecting abnormal rhythms.
For instance, AI ECG systems have achieved 99 per cent accuracy in classifying patients into normal, abnormal, and life-threatening categories. Additionally, AI has proven effective in identifying severe valvular heart disease, cardiomyopathy, and patients at risk of atrial fibrillation, an irregular rhythm linked to increased stroke risk. These developments promise earlier disease detection, improved patient triage and timely preventive measures.
Real-time heart assessment
For patients with known heart disease, shortness of breath or abnormal ECG results, echocardiography, a non-invasive ultrasound imaging technique, is used to evaluate heart size and function. AI applications in echocardiography have demonstrated that machine learning algorithms can accurately assess heart function and detect abnormal muscle motion with a level of precision comparable to that of experienced cardiologists.
A NEWSLETTER FOR YOU
Friday, 2 pm
Lifestyle
Our picks of the latest dining, travel and leisure options to treat yourself.
Studies show that AI can automate measurements in 98 per cent of echocardiogram images, with an average analysis time of about eight seconds per patient. A pivotal study published in Circulation in 2018 revealed that AI could perform fully automated echocardiogram interpretation, often with greater accuracy than manual measurements. This capability enhances resource utilisation and productivity in echocardiography.
Preventing heart attacks
Coronary artery disease is a leading cause of mortality in many developed countries. Over the past 20 years, coronary computed tomography angiography (CCTA) – a non-invasive 3D scan of the heart arteries – has become crucial for evaluating CAD severity and prognosis. Coronary artery calcium scoring (CACS), which quantifies calcium in the heart arteries, helps estimate CAD risk with minimal radiation exposure.
Machine learning-based risk stratification has shown superior performance compared to traditional models in predicting heart artery blockage. A study involving 85,945 asymptomatic participants who underwent CACS demonstrated the superiority of machine-learning algorithms over conventional risk models.
SEE ALSO
In a five-year study by Motwani et al, published in the European Heart Journal in 2017, researchers investigated the effectiveness of combining machine learning with CCTA in predicting cardiac events. The study involved 10,030 patients with suspected coronary artery disease who underwent clinically indicated CCTA.
The study revealed that machine-learning models integrating both clinical data and CCTA results significantly improved predictions of five-year mortality compared to using clinical metrics or CCTA alone. This enhancement is particularly notable in detecting unstable non-obstructive cholesterol plaques in the arterial walls – plaques that traditional non-invasive tests often miss but are associated with acute cardiac events.
By leveraging AI to analyse CCTA data more effectively, cardiologists can gain a clearer understanding of coronary artery disease severity and better predict future cardiac events. This advancement enables earlier identification of high-risk patients and supports more timely and precise interventions, ultimately aiming to reduce cardiac mortality.
Assessing heart damage
Cardiac magnetic resonance (CMR) imaging uses a strong magnetic field to produce detailed images of the heart and assess its function. Integrating AI into CMR has led to significant advancements. Machine-learning methods have achieved precision comparable to human analysis but at speeds 186 times faster.
A ground-breaking development, virtual native enhancement (VNE), generates images indicating scar tissue without the need for contrast agents, and is beneficial for patients with kidney impairment or contrast allergies. This AI application in CMR eliminates the need for X-ray radiation and contrast agents in certain studies, making heart function assessment safer and more accessible.
Diagnosis of heart failure
AI is making strides in predicting and managing heart failure. Heart failure affects 1 to 2 per cent of the adult population in developed countries and more than 10 per cent of those aged 70 and older. Early diagnosis and prediction are crucial for effective treatment and improving patient life expectancy. Machine-learning applications enhance early detection, classification, severity estimation and prediction of adverse events, including early re-hospitalisation.
A notable study by Aljaaf et al, presented at the Third International Conference on Technological Advances in Electrical, Electronics and Computer Engineering in 2015, showcased an AI-powered multi-level risk assessment model for heart failure. This model stratified risk into five categories (no risk, low, moderate, high and extremely high) and achieved a sensitivity of 86.5 per cent and specificity of 95.5 per cent.
Cardiac resynchronisation therapy (CRT), which is a specialised pacemaker implantation, is a crucial treatment for symptomatic heart failure with specific criteria. While these criteria generally indicate potential benefit from CRT, around 30 per cent of eligible patients do not experience significant clinical improvement.
Thus, predicting a patient’s response to CRT before implantation is essential for informed decision-making. AI and machine learning can enhance predictive accuracy, reduce unnecessary CRT procedures, minimise medical risks and lower hospital costs. Improved predictions also enable more effective community-based follow-up care, reducing hospital admissions and optimising resource use.
Other innovations
AI continues to drive innovation in cardiology. For example, an AI algorithm for heart murmur detection has used a dataset of 3,180 heart sound recordings to improve screening for valvular and congenital heart diseases. This system achieved a sensitivity of 93 per cent, specificity of 81 per cent and accuracy of 88 per cent. Another AI system detects coronary artery disease through facial photographs and identifies asymptomatic atrial fibrillation using contactless photoplethysmography, with 90 per cent accuracy in 30-second recordings and 97.1 per cent accuracy in 10-minute recordings.
The future of AI in cardiology
In an ideal scenario, AI-powered medical devices will enable cardiologists to diagnose heart disease earlier, accurately stratify risk and implement timely preventive measures. Despite significant advancements in medicine, reductions in heart-related deaths have plateaued in recent years. AI has the potential to drive the next major revolution in reducing cardiovascular mortality.
This article is part of a monthly series on health and well-being, produced in collaboration with Royal Healthcare
Copyright SPH Media. All rights reserved.
