AI Echocardiographic Screening of Cardiac Amyloidosis

Summary

Recent advances in machine learning and image processing techniques have shown that machine learning models can identify features unrecognized by human experts and accurately assess common measurements made in clinical practice. Echocardiography is the most common form of cardiac imaging and is routinely and frequently used for diagnosis. However, there is often subjectivity and heterogeneity in interpretation. Artificial intelligence (AI)'s ability for precision measurement and detection is important in both disease screening as well as diagnosis of cardiovascular disease.

Cardiac amyloidosis (CA) is a rare, underdiagnosed disease with targeted therapies that reduce morbidity and increase life expectancy. However, CA is frequently overlooked and confused with heart failure with preserved ejection fraction. Some estimates suggest that CA can be as prevalence as 1% in a general population, with even higher prevalence in patients with left ventricular hypertrophy, heart failure, and other cardiac symptoms that might prompt echocardiography.

AI guided disease screening workflows have been proposed for rare diseases such as cardiac amyloidosis and other diseases with relatively low prevalence but significant human impact with targeted therapies when detected early. This is an area particularly suitable for AI as there are multiple mimics where diseases like hypertrophic cardiomyopathy, cardiac amyloidosis, aortic stenosis, and other phenotypes might visually be similar but can be distinguished by AI algorithms. The investigators have developed an algorithm, termed EchoNet-LVH, to identify cardiac hypertrophy and identify patients who would benefit from additional screening for cardiac amyloidosis.

Conditions

• Cardiac Amyloidosis

Interventions

DIAGNOSTIC_TEST

EchoNet-LVH Assessment

The AI algorithm is previously described (Duffy et al. JAMA Cardiology 2022) and will remain unchanged throughout the course of the study. A pre-determined threshold based on prior experiments and analysis has been decided prior to the study. From each site, approximately 100,000 echocardiogram studies will be reviewed by EchoNet-LVH for approximately 500 patients to be flagged.

Sponsors & Collaborators

• Palo Alto Veteran Affairs Hospital

  collaborator UNKNOWN

• Providence Heart & Vascular Institute

  collaborator OTHER

• Northwestern Medicine

  collaborator OTHER

• Cedars-Sinai Medical Center

  lead OTHER

Principal Investigators

• Lily Stern, MD · Cedars-Sinai Medical Center

Study Design

Allocation

NA

Purpose

DIAGNOSTIC

Masking

NONE

Model

SINGLE_GROUP

Eligibility

Min Age

22 Years

Sex

ALL

Healthy Volunteers

No

Timeline & Regulatory

Start

2024-10-28

Primary Completion

2025-11-01

Completion

2026-11-01

FDA Device

Yes

Countries

• United States

Study Locations