Conference Coverage

AI Echocardiography Detects Cardiac Amyloidosis 13 Months Earlier Than Human Readers

Edited by:
Ashton L. Stahl

Key Highlights

  • AI-assisted echocardiography identified cardiac amyloidosis a mean of 391 days earlier than human-generated reports (P < .001).
  • Among 51 patients included in the primary analysis, all had at least one AI-positive echocardiogram, compared with 32 patients with human-positive reports.
  • In patients with both AI-positive and human-positive findings, AI detected cardiac amyloidosis first in 21 cases and on the same echocardiogram in 11 cases.
  • More than one-third of patients with AI-positive studies had no pre-diagnostic echocardiography report raising concern for cardiac amyloidosis.

Artificial intelligence (AI) applied to echocardiography identified cardiac amyloidosis approximately 13 months earlier than human readers among patients with confirmed transthyretin amyloid cardiomyopathy (ATTR-CM), according to a retrospective cohort study scheduled for presentation at the 2026 American Heart Association's (AHA) Scientific Sessions.

The findings suggest that AI-assisted echocardiography may identify features of cardiac amyloidosis that are not recognized in conventional clinical reports, potentially supporting earlier diagnosis in patients undergoing cardiac imaging.

Investigators retrospectively identified patients diagnosed with ATTR-CM between 2016 and 2022, confirmed through either a positive cardiac pyrophosphate (PYP) scan without light-chain amyloidosis or a positive endomyocardial biopsy. Eligible patients had undergone at least 2 echocardiograms before their definitive diagnosis.

Researchers analyzed pre-diagnostic echocardiograms using the FDA-cleared software EchoGo Amyloidosis and compared AI-generated findings with human interpretations. Human reports were considered positive if they contained terminology associated with cardiac amyloidosis, including "amyloid," "amyloidosis," "apical sparing," "cherry on top," or "infiltrative."

The investigators compared the timing of initial AI-positive and human-positive echocardiograms relative to definitive ATTR-CM diagnosis. They also examined the timing of initial heart failure (HF) diagnoses and loop diuretic prescriptions.

Study Findings

Of 58 patients meeting eligibility criteria, 51 patients with 160 pre-diagnostic echocardiograms had at least 1 AI-positive or human-positive study and were included in the primary analysis.

The mean age at diagnosis was 78.5 years, 11.2% were female, and 25.5% identified as Black or African American. Patients underwent a median of 3 echocardiograms before diagnosis.

All 51 patients had at least 1 AI-positive echocardiogram, whereas 32 had at least 1 human-positive report. Among patients with both AI-positive and human-positive studies, AI identified cardiac amyloidosis before human readers in 21 cases and on the same echocardiogram in 11 cases. No human-positive report preceded an AI-positive finding.

AI first detected cardiac amyloidosis a mean of 509 days before definitive ATTR-CM diagnosis, compared with 162 days for human readers. The first AI-positive echocardiogram preceded the first human-positive report by a mean of 391 days (SD, 505; P < .001).

More than one-third of patients with AI-positive studies had no pre-diagnostic echocardiography report indicating concern for cardiac amyloidosis.

Investigators also identified evidence of HF or diuretic use preceding ATTR-CM diagnosis. The first documented HF diagnosis occurred a mean of 529 days before definitive diagnosis, while the first loop diuretic prescription occurred a mean of 1,097 days before diagnosis.

Clinical Implications

According to the study authors, AI-assisted echocardiography may add clinical value by identifying subtle features of cardiac amyloidosis that human readers do not report.

The researchers suggested that applying AI to clinically indicated echocardiograms could help facilitate earlier recognition of ATTR-CM among patients whose imaging findings might otherwise go unrecognized.

Expert Commentary

“AI detected CA on echo approximately 13 months earlier than human readers and in cases not identified by human interpretation. Notably, among patients with AI-positive studies, more than one-third had no pre-diagnostic echo report that raised concern for CA,” the researchers concluded.

Reference

Sollee J, Easton N, Khan SS, et al. Artificial intelligence detects cardiac amyloidosis on echocardiography earlier than human readers. Abstract Sa1111. Presented at: American Heart Association Scientific Sessions 2026; November 7, 2026.