AI models for cardiac amyloidosis could make a world of difference

 

With new drugs recently made available to treat cardiac amyloidosis, there has been a rapid rise in interest to improve the diagnosis and treatment of these patients. Signs of amyloidosis are often found during echocardiograms, but early disease, when outcomes with drugs are optimal, is subtle and not easy to detect. And then more advanced imaging like nuclear scans or MRI is required to determine which subtype it is, since treatment varies by subtype. Artificial intelligence (AI) is expected to play a big role in better detecting and classifying the disease in the near future.

"We know that delays in diagnosis are very common in cardiac amyloidosis. Many patients may be seen for years by providers and may see multiple providers before a diagnosis is made. And we also know that when treatment is initiated at the latest stages of the disease, that life-prolonging effect is attenuated. So that really highlights the need for echo to be accurate and to detect the disease in the earlier stages when the treatment can be most effective. So that's really what our goal is with these deep learning algorithms," explained Jeremy Slivnick, MD, assistant professor of medicine and an advanced cardiac imager at the University of Chicago.

Slivnick has been involved in developing an echocardiography algorithm for enhanced cardiac amyloidosis detection. He spoke about progress in that area the the American Society of Echocardiography (ASE) 2023 meeting. He also stopped to chat with Cardiovascular Business for an in-depth look at this topic. 

Slivnick was also awarded the 2023 Arthur E. Weyman Young Investigator’s Award at ASE 2023 for his AI research.

Challenges in diagnosing cardiac amyloid

Cardiac amyloidosis, a condition where abnormal proteins accumulate in the heart tissue, has been gaining attention due to the availability of drugs for its treatment. While echo is used as the front-line imaging modality for detection, it is not as good at detection as PET scans or cardiac MR, but is often the gateway to accessing these more advanced imaging modalities. At focused ASE sessions on how to detect cardiac amyloidosis, several presenters stressed early disease can often be missed because its subtle signs in the images might be overlooked. 

Slivnick's research focuses on utilizing deep learning analysis of echocardiographic images for automated classification of regional wall motion, which can be used for several types of cardiomyopathy, including cardiac amyloidosis and hypertrophic cardiomyopathy. The AI can pull additional information from the imaging data to help boost detection and diagnostic confidence.

One of the challenges in cardiac amyloidosis lies in distinguishing between its subtypes, as the therapeutic approach varies. Slivnick revealed that the AI algorithm they designed aims to detect cardiac amyloidosis irrespective of subtypes, which include three main types of ATTR, light chain AL and wild-type. He said the intention of the AI is to alert clinicians to the need for further testing.

"I think that many patients who have suspicion for cardiac amyloidosis are going to get an echo prior to advanced imaging. And our goal is not so much necessarily to establish a diagnosis of cardiac amyloidosis, but really to use the echo to maximally extract all the information that we can out of it and to identify patients at the earliest stages where we can initiate treatment and potentially life prolonging therapies," he said.

Development of the AI model for better amyloid detection 

The AI model developed by Slivnick and his team has undergone testing on 700 amyloidosis patients from around the world, including expert centers in the United States, South America and Japan. Control groups, including patients with hypertrophic left ventricles due to other causes, were also included in the study.

Their WASE 3 study was developed as part of the World Alliance Societies of Echocardiography (WASE) Normal Values Study. The main goal of the study group is to determine if normal chamber quantification values vary across countries, geographical regions and cultures, because current echocardiographic reference values that define what is normal in an echo exam are mostly based on observations of white patients from the United States and Europe.

"We know from some of the classical features that we have been taught to look at, which would be things like impairment in global longitudinal strain or increases in left ventricular wall thickness, that although they were initially looked very accurate in a lot of earlier patient cohorts, that the accuracy is likely lower in more modern cohorts. And that may be in part due to our increased awareness of the disease and screening of patients that may have less advanced illness," Slivnick explained. He said that is where AI can help.

Looking ahead for how AI will aid detection of many diseases via echo

Slivnick highlighted the broader potential of this type of AI in echocardiography. While the current study has provided a foundation for AI in cardiac amyloidosis detection, he mentioned ongoing efforts to gather more data for control groups and potential collaboration on a larger scale. He said the medical community eagerly anticipates further developments in this promising field, which could revolutionize the early diagnosis and management of cardiac amyloidosis.

"AI offers us the potential to help raise our awareness of different diseases," Slivnick said. "In the future, we could be looking at a situation where AI raises suspicion for various diseases, and expert echocardiographers work collaboratively with AI to make accurate diagnoses."

Find more cardiac amyloidosis news and video

Dave Fornell is a digital editor with Cardiovascular Business and Radiology Business magazines. He has been covering healthcare for more than 16 years.

Dave Fornell has covered healthcare for more than 17 years, with a focus in cardiology and radiology. Fornell is a 5-time winner of a Jesse H. Neal Award, the most prestigious editorial honors in the field of specialized journalism. The wins included best technical content, best use of social media and best COVID-19 coverage. Fornell was also a three-time Neal finalist for best range of work by a single author. He produces more than 100 editorial videos each year, most of them interviews with key opinion leaders in medicine. He also writes technical articles, covers key trends, conducts video hospital site visits, and is very involved with social media. E-mail: dfornell@innovatehealthcare.com

Around the web

Ron Blankstein, MD, professor of radiology, Harvard Medical School, explains the use of artificial intelligence to detect heart disease in non-cardiac CT exams.

Eleven medical societies have signed on to a consensus statement aimed at standardizing imaging for suspected cardiovascular infections.

Kate Hanneman, MD, explains why many vendors and hospitals want to lower radiology's impact on the environment. "Taking steps to reduce the carbon footprint in healthcare isn’t just an opportunity," she said. "It’s also a responsibility."