Find care now
If you are experiencing a medical emergency, please call 911 or seek care at an emergency room.
MedStar Health Research Institute led an international research team exploring how adding demographic and clinical context to echocardiographic AI screening improves detection so patients can access life-saving treatment sooner.
An international research study led by MedStar Health Research Institute could help some patients with cardiac amyloidosis access exciting new therapies early in the disease process, when life-saving treatment is most effective.
Published in Circulation: Cardiovascular Imaging, researchers found that providing additional demographic and clinical context to a sophisticated AI algorithm examining cardiac images unlocks a more accurate path to detection. This tool could help doctors diagnose more patients with cardiac amyloidosis (CA) sooner, improving their chances of survival.
CA is a complex, multi-system disorder. It occurs when abnormal proteins misfold and collect into tough, stiff structures called amyloid fibrils. These cement-like fibrils travel through the bloodstream, depositing directly into nerves and organs, such as the heart. This makes it difficult for the heart to relax and pump blood efficiently, eventually leading to a life-threatening condition – heart failure.
The subtle symptoms of CA can mirror general cardiovascular problems, so patients can sometimes face a long, trying path to diagnosis that delays treatment. Classic warning signs such as these can easily be mistaken for age-related heart conditions:
-
Progressive shortness of breath
-
Ongoing fatigue
-
Sudden inability to tolerate exercise
-
Noticeable swelling in the legs, feet, or abdomen
CA can also cause irregular heart rhythms, so it can be mistaken for other conditions such as atrial fibrillation (AFib).
For many patients, cardiac amyloidosis can be difficult to diagnose because its symptoms often resemble other heart conditions. As a result, patients may spend months or even years seeking answers before receiving the correct diagnosis. Researchers hope tools like this one could shorten that journey and help patients begin treatment sooner.
Today’s advanced treatments include breakthrough therapies such as protein stabilizers and silencers, targeted heart failure care, and heart rhythm control that can add years or decades to a patient’s life, especially when CA is accurately diagnosed early in its progression.
Enhancing screening outcomes with AI plus clinical data.
The study mobilized a community of researchers from top clinical sites and academic medical centers in the U.S., Argentina, Brazil, and Japan. Researchers analyzed more than 1,000 patient records to evaluate how well the AI tool could identify cardiac amyloidosis.
Our primary objective aimed to overcome a critical limitation in an automated AI model published about a year before. That tool was designed to analyze imaging data from standard transthoracic echocardiogram ultrasound images.
While it could identify structural changes in the heart, this model left a screening “gray area.” In about 1 out of 10 cases, the tool couldn’t deliver a definitive answer, leaving patients and providers in limbo.
To improve the tool's ability to identify cardiac amyloidosis, our team expanded the information it could analyze. Instead of relying only on heart ultrasound images, the enhanced AI echo-clinical model incorporated additional patient information commonly available in the medical record, including:
-
Heart ultrasound images showing the structure and thickness of the heart muscle
-
Basic patient information, such as age, height, and weight
-
Blood and urine test results, including measures of kidney function
-
Electrocardiogram results that evaluate the heart's electrical activity
-
Specialized blood tests that help distinguish between different types of amyloidosis
By combining multiple sources of information, the AI was able to develop a more complete picture of each patient's health and identify those at higher risk for cardiac amyloidosis.
Eliminating the detection gray area.
The results of the research study show significant improvements in our ability to precisely screen for CA. Integrating clinical and laboratory data with deep-learning imaging analysis, the AI echo-clinical model eliminated that 10% gray zone. Every patient record the system analyzes is clearly flagged with an accurate risk level, removing ambiguity and delay.
The research study also found the new model was even more accurate than the prior version. Incorporating real-world information boosted the tool’s overall sensitivity from 76% up to 93%, while lifting overall accuracy from 80% to 90%.
We wanted to be sure the tool could be useful in the real world, where clinicians are often tasked with differentiating CA from other conditions with similar symptoms. Instead of comparing patients only to entirely healthy controls, we challenged the algorithm to tell the difference between CA and look-alike conditions.
AI-ECM maintained its high accuracy while filtering out:
-
Patients with advanced, severe aortic valve stenosis
-
Those with chronic, long-term high blood pressure (hypertension)
-
Patients with systemic amyloidosis outside of the cardiovascular system
-
Healthy control subjects with standard age-related cardiac changes
Why sensitivity rules in screening.
When screening for a serious disease, it's important to identify as many potential cases as possible. Missing a diagnosis could delay treatment. That's why researchers designed the tool to cast a wide net and alert doctors whenever cardiac amyloidosis may be present.
In this way, the AI echo-clinical model is a bit like an airport metal detector, a very sensitive device that signals an alert when any trace of dense metal passes through. It’s okay if it sometimes flags a belt buckle, known as a false positive, because the most important thing is to make sure no real security threats get through.
The AI echo-clinical model works on the same principle. It casts a broad net to make sure no patient with CA slips through, giving doctors a reliable early warning. Once the tool identifies suspected CA, providers can confidently do more tests to confirm a diagnosis.
Related: Read “New Cardiac Amyloidosis Treatments Improve Quality and Length of Life.”
Next steps: Bringing AI to the bedside.
Despite these exciting results, the AI echo-clinical model is not yet approved for day-to-day clinical use. Next, we’ll validate the tool, testing the algorithm against independent, diverse groups of patients to make sure it works for everyone.
Our ultimate vision is to build a seamless, automated background workflow that can be built into mainstream EHR systems. When it’s fully integrated, the AI could run securely behind the scenes and help detect CA early by monitoring imaging and cross-referencing with lab results and demographics to flag patients at high risk.
Related: Read “Study Reveals Exciting New Treatment for Cardiac Amyloidosis with Cardiomyopathy.”
Engineering the future of cardiovascular medicine.
The international collaboration that powered this exciting research was led by MHRI researchers who are focused on innovations in cardiovascular research and care. This breakthrough highlights our commitment to providing exceptional care for patients while simultaneously investing in advances that will protect the patients of tomorrow.
