Is it a heart attack or something else? How artificial intelligence can support diagnostics

Chest pain, shortness of breath, a brief loss of consciousness – warning signs that suggest a heart attack. But it might also be Takotsubo syndrome, also known as stress cardiomyopathy or broken heart syndrome with symptoms that resemble a heart attack. Yet it is of utmost importance to differentiate between the two conditions to initiate the right treatment. In a collaborative project between the University Hospital Zurich and ETH Zurich in Switzerland, a team of researchers explored how artificial intelligence (AI) and machine learning could help distinguish between the two cardiovascular conditions.

In this MEDICA-tradefair.com interview, resident physician Davide Di Vece explains how AI could assist in rendering a differential diagnosis based on echocardiography (an ultrasound of the heart).

Mr. Di Vece, why is it important to differentiate between the symptoms of a “typical” heart attack and those of a so-called broken heart syndrome as it relates to subsequent treatment steps? What makes this distinction so challenging?

Davide Di Vece: The clinical presentation of Takotsubo syndrome (TTS) resembles that of a patient with an acute myocardial infarction (heart attack). In both cases, the patient initially presents chest pain, shortness of breath and sometimes syncope. Since TTS is often "triggered" by emotional or physical events, the medical history may corroborate clinical suspicion of Takotsubo syndrome in certain clinical contexts. However, you always need additional diagnostic tests to confirm the suspicion. While transthoracic echocardiography is the first-line imaging modality in patients with suspected TTS, coronary angiography including ventriculography is currently considered essential to safely rule out a heart attack.

How can AI and deep learning improve diagnostics using cardiac ultrasound in these cases?

Di Vece: Echocardiography plays a pivotal role in the diagnosis of suspected TTS in patients. The main morphologic hallmark of TTS is the presence of transient myocardial wall motion abnormalities, usually extending beyond the area of ​​a single epicardial coronary artery distribution. The identification of specific patterns of regional wall motion abnormalities can support the cardiologist in identifying TTS. It can be assumed that artificial intelligence detects new patterns to distinguish between TTS and acute myocardial infarction and enhance the diagnostic value of echocardiography as it pertains to a differential diagnosis.

The study showed that AI ​​was more accurate in the applied data sets than experienced human cardiologists. While we have still a long way to go before the technology can be used in everyday clinical practice, what would a "collaboration" between AI and cardiovascular specialists look like in the future (as it pertains to acute myocardial infarction vs. Takotsubo syndrome)?

Di Vece: The skillful use of AI by trained specialists can improve the quality of the images taken. It results in more data for processing that would otherwise been unusable. Not only does this save time, but also means a cost reduction. That being said, patients benefit the most from this because AI can become another tool to facilitate a more accurate diagnosis and subsequent personalized treatment. The latter aspect is especially interesting when it comes to more complex cases with multiple comorbidities.

Keeping all these aspects in mind, the use of AI in cardiovascular imaging seems like a tool for cardiology experts that can serve clinical intelligence, though it is no replacement or substitute for the knowledge and experience of a skilled cardiologist.

The cardiologist must augment the use of AI with other clinical elements as it pertains to the differential diagnosis between myocardial infarction and Takotsubo syndrome. Indeed, our ultimate goal is to develop a comprehensive algorithm that comprises multiple parameters such as – among others – imaging data, demographic data, ECG, and biomarkers as a non-invasive tool to detect TTS. If coronary angiography is not immediately available, a plausible determination of the pretest probability of TTS can support prognostic stratification and help in determining subsequent medical treatment.