Cardiovascular

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.

TL;DR

A visually interpretable, artificial intelligence-derived ECG algorithm enables effective screening for cardiac transthyretin amyloidosis among patients with heart failure with preserved ejection fraction, with high accuracy, strong association with disease, and prognostic implications for 3-year survival.

Key Findings

The AI-derived ECG pattern was present in 82.6% of patients with ATTR-CA compared to 10.2% with HFpEF and 6.5% with hypertrophic cardiomyopathy.

  • Difference was statistically significant (P<0.001)
  • Internal validation cohort comprised 560 patients: 149 with ATTR-CA, 318 with HFpEF, and 93 with hypertrophic cardiomyopathy
  • Standard 12-lead ECGs were analyzed blindly by 3 independent observers using a previously developed, 2-step, AI-derived, visually interpretable ECG algorithm
  • Study was multicenter, drawing from 2 European centers

The ECG algorithm demonstrated high diagnostic accuracy in the internal validation cohort.

  • Area under the curve (AUC) was 0.87 (95% CI, 0.84–0.90)
  • Sensitivity was 83% (95% CI, 76%–88%)
  • Specificity was 91% (95% CI, 88%–93%)
  • Negative predictive value was 93% (95% CI, 91%–96%)

The ECG algorithm maintained good diagnostic performance in an independent external validation cohort.

  • External cohort comprised 107 patients (72 ATTR-CA, 31 HFpEF, 4 hypertrophic cardiomyopathy)
  • AUC was 0.84 (95% CI, 0.76–0.92)
  • Sensitivity was 89% (95% CI, 78%–94%)
  • Specificity was 79% (95% CI, 63%–90%)

The presence of the ECG pattern was strongly associated with a diagnosis of ATTR-CA.

  • Odds ratio for ATTR-CA was 46 (95% CI, 27–80)
  • Association was highly statistically significant (P<0.001)
  • The algorithm is described as a '2-step, artificial intelligence-derived, visually interpretable ECG algorithm'

The AI-derived ECG pattern was associated with reduced 3-year survival.

  • Log-rank P=0.007 for difference in 3-year survival based on ECG pattern presence
  • This finding suggests the algorithm carries prognostic as well as diagnostic implications
  • Survival analysis was conducted across the study population of heart failure patients

Cardiac transthyretin amyloidosis (ATTR-CA) is frequently underdiagnosed and commonly presents as heart failure with preserved ejection fraction (HFpEF), making early identification challenging in routine practice.

  • Early identification enables access to disease-modifying therapy
  • The study authors note that 'simple, widely available screening tools are needed'
  • The algorithm's compatibility with standard ECG systems was highlighted as supporting broad clinical implementation

What This Means

This research suggests that a special pattern-recognition algorithm developed using artificial intelligence can accurately identify a serious heart condition called cardiac transthyretin amyloidosis (ATTR-CA) using a standard 12-lead ECG — the routine heart tracing test available in virtually every medical setting. ATTR-CA is a disease where abnormal proteins build up in the heart muscle and is often missed because it looks similar to other causes of heart failure. In this study of 885 heart failure patients from two European hospitals, the algorithm correctly identified the ECG pattern in about 83% of ATTR-CA patients while correctly ruling it out in about 91% of patients who did not have the condition. These results held up well when tested in an independent group of patients, with sensitivity of 89% and specificity of 79%. The study also found that patients who showed this ECG pattern had significantly worse survival over 3 years, suggesting the algorithm not only helps with diagnosis but may also flag patients at higher risk of death. The odds of having ATTR-CA were 46 times higher in patients with the ECG pattern compared to those without it. Because ATTR-CA can now be treated with disease-modifying drugs, catching it early is increasingly important — but it currently requires expensive and specialized tests like cardiac MRI or nuclear imaging. This research suggests that this AI-derived ECG tool, which is visually interpretable by trained observers rather than requiring a computer at the point of care, could serve as a practical first-line screening step to identify patients who should undergo more specialized testing for ATTR-CA. Its compatibility with standard ECG equipment means it could potentially be used widely without requiring expensive new technology or infrastructure.

Have a question about this study?

Citation

Gioia G, Seirer B, Dusik F, Rettl R, Binder C, Duca F, et al.. (2026). 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.. Journal of the American Heart Association. https://doi.org/10.1161/JAHA.126.049283