What happened
Caption Health of Brisbane, California, makes Caption Interpretation Automated Ejection Fraction Software, known as AutoEF. It applies machine learning to ultrasound clips of the heart to estimate left ventricular ejection fraction, a measure of how well the heart pumps. A convolutional neural network picks the best views, and a deep learning model, locked before validation, makes the estimate. The clinician can edit the result and must accept or reject it. A retrospective study used 186 patient studies from Minneapolis Heart Institute, Duke University and Northwestern University. Expert cardiologists' biplane measurements served as the reference. FDA reviewers report that the root mean square deviation was 7.21 for the best available view, within the acceptance criterion of 9.2%. Every combination of views and each single view also met its goal. FDA granted the De Novo request on 24 February 2023, together with a plan that sets how Caption Health may retrain and update the model. The study used stored images, so the estimates guided no care. No patient outcome was measured.
Sources
-
Regulatory document, by authors with no tie to the system.
U.S. Food and Drug Administration, Center for Devices and Radiological Health. De Novo Classification Request for Caption Interpretation Automated Ejection Fraction Software (DEN220063): decision summary. Granted 2023-02-24.
The primary endpoint for the subject device was met with RMSD EF%[95% CIs]: 7.21[6.62, 7.74] for the best available view.
Details
AI system
- Workflow
-
A convolutional neural network classified the clips and selected the best available views. The model that estimates ejection fraction was trained on apical and parasternal long-axis views and locked before validation. Expert cardiologists measured the reference values with the biplane Modified Simpson's method.
Follow-up
- Conditions for replication
-
The patient is an adult with a transthoracic echocardiogram. A qualified professional checks that the images are adequate for the measurement. The clinician reviews the estimate and accepts, edits or rejects it.
Entry details
- Sector
-
Healthcare.
- Purpose
-
Caption Health built AutoEF to estimate left ventricular ejection fraction automatically from transthoracic ultrasound images and assist clinicians in a cardiac evaluation.
Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, 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). AutoEF estimated how well the heart pumps within its error goal (AIBD-0778). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0778.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)