Federated Learning for Point-of-Care Cardiac Ultrasound (FL-POCUS)

A Prospective Multicenter Clinical-Performance Study of Federated Machine Learning for Automated Interpretation of Point-of-Care Cardiac Ultrasound

This prospective, multicenter study will evaluate a federated machine-learning system designed to analyze focused cardiac point-of-care ultrasound examinations. Federated learning allows participating clinical sites to contribute to model development while keeping raw ultrasound images and directly identifiable patient information within each site's controlled computing environment. Encrypted model updates, rather than patient images, will be transmitted for secure aggregation.

The prospective validation cohort will include approximately 3,000 adults undergoing clinically indicated focused cardiac ultrasound. Model performance will be compared with an expert interpretation of a comprehensive transthoracic echocardiogram performed within 24 hours. The primary objective is to determine how accurately the model identifies reduced left ventricular systolic function, defined as a left ventricular ejection fraction below 40%.

During the initial validation period, the investigational software will operate in silent mode. Its results will not be displayed to treating clinicians and will not be used to diagnose participants, select treatment, or replace standard clinical interpretation.

The study will also evaluate image-quality classification, cardiac-view recognition, performance across clinical sites and ultrasound systems, model calibration, processing time, cybersecurity, privacy resilience, and performance across demographic and clinical subgroups. Long-term monitoring will assess whether model performance changes as clinical populations, ultrasound equipment, acquisition practices, and software environments evolve during the 2026-2037 study period.

Study Overview

• Status: Enrolling by invitation
• Conditions: Ventricular Dysfunction, Left; Ultrasonography; Machine Learning; Deep Learning; Heart Function Tests; Echocardiography; Point-of-Care Systems; Ventricular Function, Left; Stroke Volume; Neural Networks, Computer; Sensitivity and Specificity; Image Interpretation, Computer-Assisted; Artificial Intelligence (AI); Federated Learning; Diagnosis, Computer-Assisted; ROC Curve
• Intervention / Treatment: Diagnostic test: Focused Cardiac Point-of-Care Ultrasonography; Device: FL-POCUS Federated Machine-Learning Analysis System

Detailed Description

Federated Learning for Point-of-Care Cardiac Ultrasound (FL-POCUS) is a prospective, multicenter clinical-performance study of a federated machine-learning system for focused cardiac point-of-care ultrasound. The study is intended to determine whether a diagnostic model can be developed and validated across multiple clinical environments without routinely transferring raw ultrasound images or directly identifiable participant information to a central training repository.

Participating institutions may use previously collected, locally governed ultrasound examinations during the federated model-development stage. Each institution will operate a local computing node using a common model architecture, data specification, and quality-control framework. Local model updates will be encrypted and transmitted to a secure aggregation service. The aggregated parameters will then be redistributed to participating sites for subsequent training rounds. Training events, software versions, data-quality findings, and model changes will be documented in an auditable version-control system.

Privacy protections will include access controls, secure aggregation, data-minimization procedures, cybersecurity testing, and evaluation for membership-inference and model-inversion risk. Raw ultrasound images, protected health information, consent records, participant identifiers, and authentication credentials will not be placed on a public blockchain. Any distributed ledger used by the study will be limited to document hashes, version identifiers, authorized attestations, and non-sensitive audit records.

After federated development is complete, the model will be version-locked before prospective clinical validation. Approximately 3,000 adult participants will be enrolled across at least six geographically and technically diverse clinical sites. Eligible participants will be undergoing a clinically indicated focused cardiac point-of-care ultrasound examination and will have an eligible comprehensive transthoracic echocardiogram available within 24 hours of the index examination.

The locked model will operate in silent mode. Investigational outputs will not be displayed to treating clinicians and will not influence diagnosis, treatment, patient disposition, or the decision to obtain additional testing. All clinical decisions will continue to be made through the participating institution's standard care processes.

The model will evaluate standard focused cardiac views, including parasternal long-axis, parasternal short-axis, apical four-chamber, and subcostal views when available. Investigational functions may include cardiac-view classification, image-quality assessment, identification of technically limited examinations, and detection of reduced left ventricular systolic function.

The primary outcome is the area under the receiver-operating-characteristic curve for detecting a reference-standard left ventricular ejection fraction below 40%. The proposed performance criterion is an area under the curve of at least 0.85, with the lower bound of the two-sided 95% confidence interval exceeding 0.80.

The reference standard will be established using comprehensive transthoracic echocardiography. Two qualified echocardiography readers, masked to the investigational model result, will independently review eligible reference examinations. Disagreements affecting the prespecified ejection-fraction category will be resolved by a third senior reader.

Secondary evaluations will include sensitivity, specificity, positive and negative predictive values, detection of severe systolic dysfunction, cardiac-view classification accuracy, agreement with expert image-quality assessments, nondiagnostic examination rate, calibration, processing time, site-level heterogeneity, and performance by ultrasound manufacturer and transducer type.

Prespecified subgroup analyses will examine performance by age, sex, race, ethnicity, body mass index category, clinical environment, cardiac rhythm, acquisition experience, study site, and ultrasound system. Subgroup results will be reported even when the model satisfies the overall primary performance criterion.

All attempted examinations will remain in the primary intention-to-diagnose analysis, including technically limited studies and examinations with incomplete cardiac views. A model output that cannot produce a valid diagnostic classification will be counted as a test failure in the primary analysis. An evaluable-case analysis may be performed as a secondary analysis.

The study includes a long-term performance-surveillance period extending through September 30, 2037. This period will evaluate model drift, calibration changes, equipment transitions, software updates, cybersecurity events, evolving acquisition practices, and changes in the enrolled population. Prospective data used for final validation will remain separated from model-development data unless a separately governed amendment authorizes a new model version. Any updated model will receive a new version designation and must undergo independent validation before clinical use.

Reportable study events will include unauthorized data disclosure, attempted reconstruction of participant information, incorrect linkage between examinations and participants, inadvertent display of investigational results, validation-data leakage into training, material subgroup-performance disparities, cybersecurity incidents, and significant protocol deviations.

The study will not authorize automated diagnosis or autonomous clinical management. Any later investigation in which model results are displayed to clinicians or used to influence care will require a separately approved protocol, updated risk assessment, applicable regulatory review, and independent Institutional Review Board authorization.

• Study Type: Observational
• Enrollment (Estimated): 3000

Contacts and Locations

Study Locations

• United States
  • New York
    • New York, New York, United States, 10016
      • Truway Health, Inc.

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Non-Probability Sample

Study Population

The study population will consist of adults receiving care in emergency departments, inpatient units, outpatient clinics, or other participating clinical settings who undergo a clinically indicated point-of-care cardiac ultrasound examination and a reference transthoracic echocardiogram within 24 hours. Consecutive eligible participants will be enrolled without selection based on expected left ventricular function or image quality. The cohort is intended to represent diverse clinical sites, ultrasound devices, operators, demographic groups, body compositions, cardiac rhythms, and levels of left ventricular systolic function encountered in routine care.

Description

Inclusion Criteria:

• Age 18 years or older.
• Undergoing a clinically indicated point-of-care cardiac ultrasound (POCUS) examination at a participating site.
• Reference transthoracic echocardiogram completed within 24 hours of the index POCUS examination.
• At least one cardiac ultrasound view attempted during the index examination.
• POCUS and reference echocardiography records can be linked using an authorized coded study identifier.
• Participant consent obtained or inclusion authorized under an Institutional Review Board-approved consent waiver, as applicable.

Exclusion Criteria:

• Reference transthoracic echocardiogram not completed within 24 hours of the index POCUS examination.
• Major cardiac procedure or substantial hemodynamic intervention occurring between the POCUS examination and reference echocardiogram, including cardiac surgery, cardioversion, cardiac arrest, or initiation of mechanical circulatory support.
• Ultrasound or reference data are missing, corrupted, irretrievable, or cannot be securely linked.
• Previous inclusion of the same participant in the primary validation cohort, unless repeat examinations are authorized under a prespecified longitudinal analysis.
• Declines participation when individual informed consent is required.
• Member of a population not authorized for enrollment under the applicable Institutional Review Board approval.

Study Plan

How is the study designed?

Number of groups / cohorts

1

Cohorts and Interventions

Group / Cohort

Intervention / Treatment

Prospective Silent-Mode Federated Cardiac Ultrasound Validation Cohort; Approximately 3,000 adults undergoing clinically indicated focused cardiac point-of-care ultrasonography will be included in this prospective cohort. Each participant's cardiac ultrasound examination will be evaluated by the locked Federated Learning for Point-of-Care Cardiac Ultrasound (FL-POCUS) machine-learning model and compared with a reference transthoracic echocardiogram completed within 24 hours. Investigational model outputs will remain in silent mode and will not be displayed to treating clinicians or used to direct diagnosis, treatment, patient disposition, or additional testing. All attempted examinations, including technically limited studies and examinations with incomplete views, will remain in the primary intention-to-diagnose analysis.

Diagnostic test: Focused Cardiac Point-of-Care Ultrasonography; A clinically indicated, noninvasive focused cardiac ultrasound examination performed through a point-of-care ultrasound system. Standard views may include parasternal long-axis, parasternal short-axis, apical four-chamber, and subcostal views. The examination will be evaluated for image quality, cardiac-view classification, left ventricular function, and evidence of reduced left ventricular ejection fraction.; Other Names: Point-of-Care Ultrasound; POCUS; Focused Cardiac Ultrasound; FoCUS; Cardiac Ultrasonography; Device: FL-POCUS Federated Machine-Learning Analysis System; Investigational software that analyzes focused cardiac point-of-care ultrasound examinations using a version-locked federated machine-learning model. The system evaluates cardiac-view classification, image quality, left ventricular function, and the probability of a left ventricular ejection fraction below 40%. During this observational validation study, all model outputs will remain in silent mode and will not influence clinical care. Raw ultrasound images and directly identifiable participant information will remain within each participating site's controlled environment.; Other Names: FL-POCUS; Federated Learning POCUS System; Federated Cardiac Ultrasound Model; Silent-Mode Echocardiography AI

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Diagnostic Performance for Detecting Reduced Left Ventricular Systolic Function	Area under the receiver operating characteristic curve (AUROC) of the locked federated-learning point-of-care ultrasound (POCUS) model for identifying left ventricular ejection fraction below 40%, using masked expert core-laboratory interpretation of the reference transthoracic echocardiogram as the reference standard. Analysis will be performed at the participant level with a two-sided 95% confidence interval.	Day 1 (within 24 hours after the index point-of-care ultrasound examination)

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Sensitivity and Specificity for Detecting Left Ventricular Ejection Fraction Below 40%	Participant-level sensitivity and specificity of the locked model at the prespecified decision threshold for detecting left ventricular ejection fraction below 40%, compared with masked expert core-laboratory interpretation of the reference transthoracic echocardiogram. Results will include two-sided 95% confidence intervals.	Day 1 (within 24 hours after the index point-of-care ultrasound examination)
Diagnostic Performance for Detecting Severe Left Ventricular Systolic Dysfunction	AUROC of the locked model for identifying severe left ventricular systolic dysfunction, defined as a reference left ventricular ejection fraction below 30%. Sensitivity and specificity at the prespecified decision threshold will also be reported.	Day 1 (within 24 hours after the index point-of-care ultrasound examination)
Cardiac Ultrasound View Classification Accuracy	Percentage of acquired ultrasound clips correctly classified by the model as parasternal long-axis, parasternal short-axis, apical four-chamber, subcostal, or other view, compared with blinded expert-reader classification.	Day 1 (index point-of-care ultrasound examination)
Agreement of Automated and Expert Image-Quality Classification	Agreement between model-generated and blinded expert-reader image-quality ratings. Images will be categorized as diagnostic, technically limited, or nondiagnostic. Percentage agreement and weighted kappa with a two-sided 95% confidence interval will be reported.	Day 1 (index point-of-care ultrasound examination)
Nondiagnostic Model Output Rate	Percentage of attempted participant examinations for which the locked model cannot generate a valid left ventricular systolic-function classification because of inadequate image quality, incomplete acquisition, unsupported cardiac view, or technical processing failure.	Day 1 (index point-of-care ultrasound examination)
Calibration of Predicted Reduced Left Ventricular Function Risk	Agreement between the model-predicted probability and observed occurrence of left ventricular ejection fraction below 40%. Calibration will be evaluated using the calibration intercept, calibration slope, and Brier score.	Day 1 (within 24 hours after the index point-of-care ultrasound examination)
Model Processing Time	Median elapsed time, measured in seconds, from availability of the completed ultrasound examination to generation of the locked model output. The interquartile range and 95th percentile will also be reported.	Day 1 (index point-of-care ultrasound examination)
Cross-Site and Ultrasound-Device Generalizability	AUROC for detecting left ventricular ejection fraction below 40% will be calculated separately by participating site and ultrasound-device manufacturer. Performance heterogeneity will be summarized using the range of site-specific and device-specific AUROCs and a hierarchical random-effects analysis.	From model lock through primary completion, up to 10 years.
Longitudinal Model Performance Drift	Annual change in AUROC, sensitivity, specificity, calibration slope, and nondiagnostic output rate relative to the first completed validation year. The model and decision threshold will remain locked during prospective validation unless a protocol-defined model update is separately evaluated.	Annually from model lock through primary completion, up to 10 years.

Other Outcome Measures

Outcome Measure	Measure Description	Time Frame
Federated-Learning Privacy Attack Resistance	Resistance of the federated-learning system to prespecified membership-inference and model-inversion tests. Membership-inference attack AUROC and the percentage of testing attempts producing recognizable reconstruction of source ultrasound data will be reported.	Before prospective deployment and annually through study completion, up to 11 years.

Collaborators and Investigators

• Sponsor: Truway Health, Inc.
• Investigators: Principal Investigator: Gavin Solomon, MD, Truway Health, Inc.

Publications and helpful links

General Publications

• Madani A, Arnaout R, Mofrad M, Arnaout R. Fast and accurate view classification of echocardiograms using deep learning. NPJ Digit Med. 2018;1:6. doi: 10.1038/s41746-017-0013-1. Epub 2018 Mar 21.
• Mor-Avi V, Khandheria B, Klempfner R, Cotella JI, Moreno M, Ignatowski D, Guile B, Hayes HJ, Hipke K, Kaminski A, Spiegelstein D, Avisar N, Kezurer I, Mazursky A, Handel R, Peleg Y, Avraham S, Ludomirsky A, Lang RM. Real-Time Artificial Intelligence-Based Guidance of Echocardiographic Imaging by Novices: Image Quality and Suitability for Diagnostic Interpretation and Quantitative Analysis. Circ Cardiovasc Imaging. 2023 Nov;16(11):e015569. doi: 10.1161/CIRCIMAGING.123.015569. Epub 2023 Nov 13.
• Dayan I, Roth HR, Zhong A, Harouni A, Gentili A, Abidin AZ, Liu A, Costa AB, Wood BJ, Tsai CS, Wang CH, Hsu CN, Lee CK, Ruan P, Xu D, Wu D, Huang E, Kitamura FC, Lacey G, de Antonio Corradi GC, Nino G, Shin HH, Obinata H, Ren H, Crane JC, Tetreault J, Guan J, Garrett JW, Kaggie JD, Park JG, Dreyer K, Juluru K, Kersten K, Rockenbach MABC, Linguraru MG, Haider MA, AbdelMaseeh M, Rieke N, Damasceno PF, E Silva PMC, Wang P, Xu S, Kawano S, Sriswasdi S, Park SY, Grist TM, Buch V, Jantarabenjakul W, Wang W, Tak WY, Li X, Lin X, Kwon YJ, Quraini A, Feng A, Priest AN, Turkbey B, Glicksberg B, Bizzo B, Kim BS, Tor-Diez C, Lee CC, Hsu CJ, Lin C, Lai CL, Hess CP, Compas C, Bhatia D, Oermann EK, Leibovitz E, Sasaki H, Mori H, Yang I, Sohn JH, Murthy KNK, Fu LC, de Mendonca MRF, Fralick M, Kang MK, Adil M, Gangai N, Vateekul P, Elnajjar P, Hickman S, Majumdar S, McLeod SL, Reed S, Graf S, Harmon S, Kodama T, Puthanakit T, Mazzulli T, de Lavor VL, Rakvongthai Y, Lee YR, Wen Y, Gilbert FJ, Flores MG, Li Q. Federated learning for predicting clinical outcomes in patients with COVID-19. Nat Med. 2021 Oct;27(10):1735-1743. doi: 10.1038/s41591-021-01506-3. Epub 2021 Sep 15.
• Rieke N, Hancox J, Li W, Milletari F, Roth HR, Albarqouni S, Bakas S, Galtier MN, Landman BA, Maier-Hein K, Ourselin S, Sheller M, Summers RM, Trask A, Xu D, Baust M, Cardoso MJ. The future of digital health with federated learning. NPJ Digit Med. 2020 Sep 14;3:119. doi: 10.1038/s41746-020-00323-1. eCollection 2020.
• Narang A, Bae R, Hong H, Thomas Y, Surette S, Cadieu C, Chaudhry A, Martin RP, McCarthy PM, Rubenson DS, Goldstein S, Little SH, Lang RM, Weissman NJ, Thomas JD. Utility of a Deep-Learning Algorithm to Guide Novices to Acquire Echocardiograms for Limited Diagnostic Use. JAMA Cardiol. 2021 Jun 1;6(6):624-632. doi: 10.1001/jamacardio.2021.0185.
• Ouyang D, He B, Ghorbani A, Yuan N, Ebinger J, Langlotz CP, Heidenreich PA, Harrington RA, Liang DH, Ashley EA, Zou JY. Video-based AI for beat-to-beat assessment of cardiac function. Nature. 2020 Apr;580(7802):252-256. doi: 10.1038/s41586-020-2145-8. Epub 2020 Mar 25.
• Gallant C, Bernard L, Kwok C, Wichuk S, Noga M, Punithakumar K, Hareendranathan A, Becher H, Buchanan B, Jaremko JL. AI-Augmented Point of Care Ultrasound in Intensive Care Unit Patients: Can Novices Perform a "Basic Echo" to Estimate Left Ventricular Ejection Fraction in This Acute-Care Setting? J Clin Med. 2025 Apr 23;14(9):2899. doi: 10.3390/jcm14092899.
• Alpert EA, Kwartz T, Hahn B, Abdulghani W, Nama A, Dadon Z. Artificial Intelligence in Cardiac Point-of-Care Ultrasound: A Narrative Review. Diagnostics (Basel). 2026 Jun 21;16(12):1921. doi: 10.3390/diagnostics16121921.

Helpful Links

• Official sponsor website and contact portal for information, research updates, and inquiries concerning the Federated Learning for Point-of-Care Cardiac Ultrasound (FL-POCUS) study.
• American Society of Echocardiography recommendations standardizing cardiac point-of-care ultrasound (POCUS) terminology, clinical scope, left-ventricular assessment, and research nomenclature.
• American Society of Echocardiography guidance for cardiac point-of-care ultrasound (POCUS), critical-care echocardiography, operator training, image acquisition, quality assurance, and laboratory oversight.
• Monarch Initiative phenotype reference for reduced left ventricular ejection fraction, supporting standardized biomedical terminology and computable phenotype mapping.
• Monarch Initiative phenotype reference for left ventricular systolic dysfunction, supporting standardized condition mapping across participating research sites.
• National Institute of Standards and Technology (NIST) Artificial Intelligence Risk Management Framework supporting trustworthy model validation, transparency, performance monitoring, privacy, security, and governance.
• Official Digital Imaging and Communications in Medicine (DICOM) standard supporting interoperable storage, transmission, retrieval, processing, and exchange of cardiac ultrasound studies.
• Health Level Seven Fast Healthcare Interoperability Resources (HL7 FHIR) ImagingStudy specification for standardized representation, retrieval, and controlled exchange of Digital Imaging and Communications in Medicine (DICOM) study information.

Study record dates

Study Major Dates

• Study Start (Actual): August 31, 2026
• Primary Completion (Estimated): October 1, 2036
• Study Completion (Estimated): September 30, 2037

Study Registration Dates

• First Submitted: August 27, 2026
• First Submitted That Met QC Criteria: August 31, 2026
• First Posted (Actual): September 2, 2026

Study Record Updates

• Last Update Posted (Actual): September 2, 2026
• Last Update Submitted That Met QC Criteria: August 31, 2026
• Last Verified: August 1, 2026

More Information

Terms related to this study

• Keywords: LVEF; Point-of-Care Ultrasound; Focused Cardiac Ultrasound; Left Ventricular Ejection Fraction; POCUS; Cardiac Ultrasound; FoCUS; Reduced LVEF; FL-POCUS; Federated Machine Learning; Privacy-Preserving Machine Learning; Cardiac View Classification; Ultrasound Image Quality Assessment; Computer-Assisted Echocardiography; Silent-Mode Clinical Validation; Multicenter External Validation; Machine-Learning Model Drift; Secure Model Aggregation
• Additional Relevant MeSH Terms: Cardiovascular Diseases; Pathologic Processes; Heart Diseases; Immune System Diseases; Ventricular Dysfunction; Pathological Conditions, Signs and Symptoms; Hypersensitivity; Disease; Ventricular Dysfunction, Left
• Other Study ID Numbers: TH-POCUS-FEDERATE-2026-001

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES
• IPD Plan Description: De-identified individual participant data underlying the reported study results will be made available. Shared data may include coded demographic characteristics, clinical setting, ultrasound-device category, POCUS acquisition characteristics, expert reference left ventricular ejection fraction, locked-model predictions, image-quality classifications, and analyzed outcome variables. Direct identifiers will not be shared. Raw ultrasound files will be available only through controlled access when permitted by the originating site, informed-consent terms, and applicable privacy requirements.
• IPD Sharing Time Frame: Data will become available beginning 12 months after publication of the primary study results and will remain available for 10 years.
• IPD Sharing Access Criteria: Access will be considered for qualified scientific researchers who submit a methodologically sound research proposal. Requests will undergo review by the sponsor's data-access committee and, when applicable, participating institutions. Approved researchers must execute a data-use agreement, protect confidentiality, refrain from re-identification, use the data only for the approved purpose, and comply with applicable Institutional Review Board, privacy, security, and publication requirements. Data will be provided through a controlled-access environment.
• IPD Sharing Supporting Information Type: STUDY_PROTOCOL; SAP; ICF; ANALYTIC_CODE; CSR

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: Yes
• product manufactured in and exported from the U.S.: No