What happened
Ultromics built EchoGo Heart Failure, machine learning software that reads one routine echocardiogram clip. It gives a yes-or-no reading for heart failure with preserved ejection fraction (HFpEF). Singapore's Agency for Care Effectiveness (ACE) reviewed the evidence in a brief published on 30 April 2025. ACE drew on three retrospective case-control studies and one retrospective real-world study. One of the case-control studies was a conference abstract. Sensitivity ranged from 87.8% to 89.8% and specificity from 55.6% to 86.3%. In the main study, two clinical scores left 54.6% and 60.4% of cases non-diagnostic, against 7.3% for the software. It reclassified 73.5% of cases that the HFA-PEFF score left indeterminate. ACE wrote that it appears less accurate than traditional clinical scores and that it can reclassify the cases they leave indeterminate. ACE found limited data on clinical utility and no cost-effectiveness study. Most studies were supported by Ultromics grants or had authors employed by the company. ACE notes that its brief rests on a limited literature search and does not appraise the studies.
Sources
-
Assessment report, by authors with no tie to the system.
Based on three retrospective case-control studies (one of which is a conference abstract) and one retrospective real-world study, EchoGo Heart Failure appeared to be safe and accurate, with limited clinical utility data.
Although it appears to be less accurate than traditional clinical scores, it can reclassify indeterminate cases resulting from them.
Details
AI system
- Workflow
-
The software read one apical four-chamber video clip from a routine transthoracic echocardiogram showing an ejection fraction of at least 50%. It sent a binary HFpEF classification to the physician's workstation through a cloud server within 20 minutes.
Follow-up
- Conditions for replication
-
Patients are adults over 25 years suspected of heart failure, with an ejection fraction of at least 50%. An interpreting clinician uses the result as an adjunct to standard diagnostic pathways. Cloud processing of patient data must meet data protection rules, such as Singapore's Personal Data Protection Act.
Entry details
- Sector
-
Healthcare.
- Purpose
-
Ultromics built EchoGo Heart Failure to help clinicians detect heart failure with preserved ejection fraction from a single routine echocardiogram clip.
30 more questions not yet checked
model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this record
Peter H. Vartanian. (2026). EchoGo Heart Failure reclassified most cases that scores left open (AIBD-1162). Artificial Intelligence Benefits Database, edition of 1 November 2026. https://petervartanian.xyz/aibd/records/AIBD-1162.html (archived at https://doi.org/10.7910/DVN/AADSII, version 7.0)