AI tool may help doctors spot ATTR-CM on routine heart test
Algorithm derived from ECG helps spot condition early
Written by |
An artificial intelligence (AI)-derived algorithm analyzing data from a standard, noninvasive heart test may help doctors identify transthyretin amyloid cardiomyopathy (ATTR-CM) among people with heart failure and preserved ejection fraction (HFpEF), a study suggested.
HFpEF, characterized by a heart muscle that is too stiff to relax properly between heartbeats, can be caused by ATTR-CM or other heart conditions. The algorithm looks for two patterns of heart electrical activity associated with ATTR-CM that can be visually detected on a standard electrocardiogram (ECG).
Across two separate groups of people with HFpEF, the algorithm identified most of those with ATTR-CM, while correctly ruling out those without the disease. It also performed well among people with no or mild heart failure symptoms.
The findings suggest that “this simple ECG‐based screening approach enables clinicians to recognize [ATTR-CM] earlier during routine evaluation of heart failure with preserved ejection fraction, potentially facilitating timely referral for confirmatory testing and earlier initiation of disease‐modifying therapies,” the researchers wrote.
The study, “Validation of an Artificial Intelligence–Derived ECG Algorithm for Detecting Cardiac Amyloidosis in Patients With Heart Failure With Preserved Ejection Fraction: Clinical Application and Prognostic Implications,” was published in the Journal of the American Heart Association.
Difficult to diagnose
ATTR-CM occurs when abnormal clumps of the transthyretin protein build up in the heart. Over time, these deposits make the heart muscle thicker and stiffer, leading to ATTR-CM symptoms such as an irregular heart rate, fatigue, and swelling in the legs or ankles.
The disease “is increasingly recognized as an underdiagnosed yet significant cause of heart failure with preserved ejection fraction,” the researchers wrote, “with up to 20% of patients with HFpEF showing evidence of [ATTR‐CM].”
Because other common causes of HFpEF can produce similar symptoms, ATTR-CM can be difficult to recognize, potentially delaying an ATTR-CM diagnosis and early access to ATTR-CM treatments. Confirming the condition can also require advanced imaging, genetic testing, and, in some cases, a biopsy, in which a small sample of tissue is collected for analysis under a microscope.
A team of researchers in Austria previously developed a two-step, AI-derived ECG algorithm designed to identify ATTR-CM among people with HFpEF.
They identified two distinct patterns of electrical activity associated with ATTR-CM that doctors can look for directly on a standard ECG, without specialized AI software or advanced imaging.
Now, the same team, along with researchers in Germany, set out to further validate this tool in two different populations of HFpEF patients.
They first tested the algorithm using ECG recordings from 560 people with HFpEF at the Medical University of Vienna in Austria. Three cardiologists independently reviewed the ECGs for the two patterns defined by the algorithm without knowing the participants’ diagnoses.
After excluding ECGs that did not meet quality standards or could not be reliably classified due to technical limitations, the analysis included 149 people with ATTR-CM, 313 with HFpEF alone, and 92 with hypertrophic cardiomyopathy (HCM), another condition that can cause the heart muscle to become abnormally thick.
The characteristic ECG pattern identified by the algorithm was present in 82.6% of people with ATTR-CM, compared with 10.2% of those with HFpEF and 6.5% of those with HCM.
Overall, the algorithm had a sensitivity of 83%, meaning it correctly identified 83% of people with ATTR-CM, and a specificity of 91%, meaning it correctly ruled out 91% of those without ATTR-CM. The cardiologists also showed a high level of agreement, indicating the patterns could be consistently recognized by different doctors.
The approach also performed well among people with no or only mild heart failure symptoms, correctly identifying 75% of those with ATTR-CM and ruling out 92% of those without the disease.
The researchers then tested the tool in an independent group of 62 people with ATTR-CM, 31 with HFpEF, and four with HCM who were seen at the Heart Center Leipzig in Germany.
Again, the characteristic ECG pattern was significantly more common among people with ATTR-CM than among those without (89% vs. 21%). The algorithm had a sensitivity of 89% and a specificity of 79%, supporting its performance in an independent group.
The presence of the distinctive pattern was significantly associated with a 10 to 46 higher likelihood of having ATTR-CM across populations. The test also showed signs of “moderate‐to‐strong rule‐in capacity and … robust rule‐out performance that is largely transportable across clinical settings,” the researchers wrote.
“The AI‐derived ECG algorithm offers a simple, scalable, and accurate approach to identifying [ATTR‐CM] in patients with HFpEF,” they wrote.
Still, they added, “the algorithm should be viewed as a screening or “red flag” tool rather than a diagnostic test.”
“By streamlining diagnosis and facilitating timely treatment, this tool may help improve outcomes for patients with [ATTR-CM],” the team concluded.
Leave a comment
Fill in the required fields to post. Your email address will not be published.