Patients who show symptoms that may indicate a heart condition often have an ECG made. In some cases, however, a definitive or unambiguous diagnosis cannot be made on the basis of a short-term ECG. In that case, a so-called long-term ECG can be chosen. In this case, an ECG, or rather a 'film', is made over a period of several days. Analyzing a long-term ECG is a time-consuming and intensive job due to the large amount of data. AI, as shown by a recent international study, can speed up this process and at the same time improve the accuracy of the diagnoses.
An AI model that is able to analyze long-term ECGs was tested for the international study. The aim was not only to find out whether the AI model can provide a solution to the shortage of staff, but also whether it has a positive effect on the quality of the diagnoses made. The researchers were brief about the latter. The AI tool reduced the number of missed diagnoses by a factor of 14. The study was led by Linda Johnson, associate professor of cardiovascular epidemiology at Lund University in Sweden, together with Jeff Healey, senior scientist at the Population Health Research Institute, a joint institute of McMaster University and Hamilton Health Sciences in Canada. The findings are published in Nature Medicine.
200,000 days of ECG data
The study included 14,606 individual patients who had recorded an average of 14 days of ECG, totaling more than 200,000 days of ECG data. These data were assessed by ECG technicians using standard clinical methodology. The same data were then reanalyzed using an AI algorithm ('DeepRhythmAI') developed specifically for this task by MEDICALgorithmics, Poland.
“We then randomly selected over 5,000 episodes of arrhythmia for intensive, beat-by-beat analysis by 17 panels of expert physicians (primarily cardiologists and electrophysiologists) from around the world, which yielded an extremely high-quality gold standard diagnosis against which we then compared ECG and AI algorithm interpretation,” Johnson said.
14x fewer missed diagnoses
The researchers found that long-term ECGs analyzed by the AI tool reduced the number of missed diagnoses by a factor of 14, for diagnoses such as serious arrhythmias, including complete heart block, ventricular tachycardia, and atrial fibrillation. The rate of missed diagnoses of serious arrhythmias using the AI analysis was just 0.3 percent, compared to a much higher rate of 4.4 percent for clinicians.
However, the researchers did not set out to prove that AI is as good or better than cardiologists at diagnosing specific arrhythmias. If the research shows that AI can be used successfully, it would be a major innovation that could address the global shortage of trained personnel capable of interpreting ECG monitoring over the long term.
"There is a shortage of approximately 15 million healthcare workers worldwide. Ambulatory ECGs need to be analyzed by specially trained personnel, often called ECG technicians. Staff shortages are a major bottleneck in healthcare worldwide, and at the same time, patients would benefit from more and longer ambulatory ECG recordings, not shorter ones," Johnson said.
AI diagnostics
The added value that AI tools and algorithms have for improving diagnostics and supporting (overburdened) healthcare professionals has already led to several promising solutions in recent years. For example, another recent study showed that congenital heart defects can be diagnosed more accurately and better using an AI-based computer program.
Last year, researchers from Iraq and Australia succeeded in developing an AI algorithm that can diagnose various diseases based on photos of the color of the tongue with 98 percent accuracy. Think diabetes, stroke, anemia, asthma, liver and gallbladder disease, COVID-19 and a range of vascular and gastrointestinal diseases.
