HeartSciences

Artificial intelligence is a feature layer on ECG management software whose value stands without it. What a facility can buy is MyoVista Insights, a cloud ECG worklist, waveform review, reporting and record integration system, which the company's annual report classifies as a Medical Device Data System.

Its AI-ECG Algorithm Marketplace displays results from third party algorithms inside that workflow, and the one available is Bunkerhill Health's cleared ECG-EF model for reduced ejection fraction. HeartSciences' own AI-ECG algorithms, licensed in part from Mount Sinai, are listed as pre-FDA or in development, and the site footer states they are not commercially available in the United States.

The oversight structure is described and the platform's own part in it is missing. A clinician reviews each ECG with any AI-ECG result displayed alongside it, then interprets, annotates and signs the report, with an audit trail on every report; the one algorithm available, Bunkerhill Health's ECG-EF model, is assistive under its own clearance. The company is explicit that its own AI-ECG algorithms are not available in the United States, rather than presenting pipeline as product.

What is not described is how the platform presents a result (a flag, a probability, a category), whether confidence is shown, what the confirm-and-next review flow advertised on the home page does when an AI result is abnormal, and what a site is told when a listed model is updated or withdrawn. Those choices shape how a clinician weighs the output, and they belong to the platform rather than to the developer.

The supplier of the model in the product is named, without the version and update discipline behind it. The AI a customer receives today is Bunkerhill Health's ECG-EF model, cleared to detect left ventricular ejection fraction at or below 40 percent from a routine 12-lead ECG, delivered through the MyoVista Insights marketplace since version 1.3 in June 2026. The platform itself has moved to version 1.4, and nothing published says which version of the Bunkerhill model is running or how a site learns that a listed model has changed.

The company's own pipeline is described in more depth than what it sells: its publications include a vision transformer foundation model for ECG (HeartBEiT, npj Digital Medicine 2023) and task models for ejection fraction, diastolic function, valve disease and pulmonary embolism, none of them available through the platform.

The model provider is named, and the rest of the chain is partial. The algorithm results a customer sees come from Bunkerhill Health's ECG-EF model, and the platform and its AI-ECG processing run on AWS, which the annual report names as the primary cloud provider. The annual report also describes outsourced IT and cybersecurity providers, unnamed.

What is not published is the path of a patient's ECG between those parties: whether the recording or derived data reaches Bunkerhill, what retention and training terms govern it there, and whether further developers added to the marketplace will be listed as they arrive. No subprocessor list is published.

Named customers, and the published research is about algorithms the platform does not yet deliver. North Shore Health, a community hospital in Cook County, Minnesota, signed in May 2026 as the first hospital deployment in the Cibolo Rural Health Networks, which endorsed the platform for its member hospitals in January 2026; the University of the West of England adopted it for an ECG education center; and the company reports its first US commercial agreements and a selection by a European reference center it does not name. The company reported no meaningful revenue for the fiscal year ended 30 April 2026 or the quarter to 31 July 2026. No outcome data from any deployment is published.

The clinical evidence page lists more than 20 peer reviewed papers, largely Mount Sinai work on ECG models for ejection fraction, right ventricular function, valve disease and diastolic function (for example, right ventricular size and function, Journal of the American Heart Association 2024, AUC 0.86), and says algorithms surfaced through MyoVista Insights are validated this way. None of those models is available through the platform; the algorithm it delivers is Bunkerhill Health's, whose evidence sits on Bunkerhill's record.

General assurances about a platform that holds identified ECG studies. The site describes HIPAA-compliant AWS infrastructure, a HIPAA-aligned audit trail and role based access on every record, and the annual report says the company expects to collect, process and retain patient information as a business associate. The only privacy policy published covers the website's contact and demo forms.

Nothing states how long studies and reports are kept, whether customer ECGs are used to develop the company's own algorithms (the annual report describes a Mount Sinai arrangement that includes de-identified data access for research), or what is returned or deleted when a contract ends. The marketplace adds a flow the company leaves undescribed. When a third party algorithm runs on a patient's ECG, there is no word on whether the recording goes to the developer or is processed inside the platform, or under what terms. Safety engineering for the algorithms displayed is not published.

Compliance is claimed and the documents behind it are not published. The site footer reads HIPAA Compliant, the platform is described as running on HIPAA-compliant AWS infrastructure with a HIPAA-aligned audit trail, and the annual report says the company expects to handle patient information as a business associate of covered entities. The only privacy policy published covers the website's contact and demo forms.

No business associate agreement, template or summary of its terms is published, and nothing states how subcontractors are bound. The marketplace makes that last point specific: when a third party algorithm runs on a patient's ECG, the developer may also handle protected health information, and whether that runs under a subcontract from HeartSciences or a direct agreement with the hospital decides who notifies whom after a breach.

An outside check is described without its type or cadence, and no attestation. The annual report (Item 1C, filed July 2026) says the company uses manual processes, software and automated tools, and third party assessments in its cybersecurity program, and that an outside provider designs and manages its IT infrastructure and security program with senior management. It also states that, as a smaller reporting company, it has no employee with significant professional IT management experience, and that the audit committee intends to oversee cyber risk.

The site describes HIPAA-compliant AWS infrastructure, which is the host's position rather than the company's. No SOC 2, HITRUST or ISO/IEC 27001 attestation for HeartSciences is named, and the ISO 13485 claim in the site footer covers quality management rather than information security.

No device claim is made for what is sold, and the company scopes its own AI out of the offer. The annual report classifies MyoVista Insights as a Medical Device Data System, exempt from 510(k), and the platform page says the marketplace shows results only from algorithms with regulatory clearance in the customer's region; the one listed is Bunkerhill Health's cleared ECG-EF model, whose clearance is Bunkerhill's. HeartSciences' own algorithms are labeled pre-FDA or in development, and the site footer states they are not commercially available in the United States.

The MyoVista wavECG device, a 12-lead ECG with the conventional Glasgow interpretation algorithm and no AI, has been under 510(k) review since December 2025. The company's September 2026 results say it does not intend to commit significant further resources to commercializing the device, that the AI algorithm needed for its commercial case would require significant further development and regulatory work, and that no probable timeline can be set. One page reads against the footer: the clinical evidence page says algorithms surfaced through MyoVista Insights are validated through the company's published studies, when the algorithm surfaced is a third party's. The site footer also claims ISO 13485 certification and a CE mark without saying which product they cover.

One listing rule is stated, and no evaluation stands behind the models the platform delivers. The MyoVista Insights page says the marketplace displays results only from enabled AI-ECG algorithms with regulatory clearance in the customer's region, and the company's clinical evidence page describes the method its own research program follows (imaging reference standards, validation at independent centers, AUC and confidence intervals per cohort). The annual report lists model drift and bias among the risks its AI-ECG algorithms carry.

Nothing is published on how the platform checks a listed algorithm's performance across groups, monitors it after deployment, or tells a site when a listed model changes. ECG interpretation varies by sex, age and body habitus, and the platform is sold to community and critical access hospitals and primary care practices that have little capacity to test a model themselves, so the marketplace's own standard is what those buyers rely on. Regulatory clearance is the only condition stated.

Review and sign off controls exist, and nothing stands behind an AI result. The platform puts third party AI-ECG results beside the study in the clinician's review, with a digital signature workflow and an audit trail on every report, so a clinician reviews before a report is finalized and the record shows who signed what. No error rate is published for the algorithm the marketplace delivers (that figure sits with its developer, Bunkerhill Health), and no guarantee, indemnity or correction route runs from HeartSciences to a customer or patient when a displayed result is wrong.

Who answers for a marketplace result is the open question: the platform displays it, the developer owns the clearance, and neither the customer terms nor the allocation of responsibility between them is published.

A named record system with a vendor program behind it, and the standards spelled out. MyoVista Insights received Epic Toolbox designation for the ECG Management System category in March 2026, meaning it meets Epic's recommended integration practices for that category, and the site links its Epic Showroom listing. The integration is described specifically: HL7 v2 order and result messaging for solicited and unsolicited ECG orders, discrete result filing, HL7 FHIR R4, SMART on FHIR app launch with OAuth 2.0 single sign-on, billing messages, and PDF delivery where the record system supports it, without a separate integration engine.

What is missing is production evidence. The company names its first commercial hospital (North Shore Health) and describes early adopter deployments, but no site is named as live on Epic or any other record system, and no conformance statement or supported version list is published.

A single cloud option on a named provider, with location implied rather than committed. MyoVista Insights is a zero footprint, auto-scaling service on AWS, and the annual report says the company relies primarily on AWS for the platform and its AI-ECG processing, naming regional outages and changes in data residency offerings among the risks of that concentration. ECG files arrive from existing carts from more than 30 manufacturers, including by secure file share ingestion, so no hardware replacement is needed.

No AWS region, residency commitment, tenancy model, backup or recovery posture, or subprocessor list is published, and there is no on premises option, so every study leaves the institution for the vendor's cloud.

No price is published and the posture is discoverable. The annual report states the company expects to earn installation fees, software as a service usage fees and fees for AI-ECG algorithms delivered through the marketplace, including third party algorithms, so a buyer knows what drives the cost before calling. No figure for any of them is published, and the demo request form is the route in.

The company cites Medicare reimbursement for assistive AI-ECG assessments under the Hospital Outpatient Prospective Payment System, APC 5734, at about $136 per assessment for 2026. That is what a hospital outpatient department may be paid, not what the platform costs, and it applies only where a cleared assistive algorithm is used.

Coverage is named, and deployments stand behind part of it. MyoVista Insights handles resting 12-lead ECGs from more than 30 manufacturers' devices, for cardiology and primary care, and is positioned for facilities from critical access and community hospitals to large health systems. The deployments named so far sit at the small end: a rural community hospital in Minnesota through the Cibolo network and a university ECG education center in England, with early adopter sites unnamed.

The AI function has narrower coverage than the platform: one third party model for reduced ejection fraction in adults, validated by its developer under its own clearance. The company's own algorithms for other findings are in development and not available.