ECG-AI Explained: What It Means for Healthcare Practices

ECG-AI is the category of FDA-regulated software-as-a-medical-device that analyzes standard 12-lead electrocardiogram data using machine learning algorithms to detect conditions that would previously have required specialized testing or clinical pattern recognition from a subspecialist. On April 8, 2026, Anumana received the first and only FDA clearance for an ECG-AI algorithm targeting cardiac amyloidosis — a serious but under-diagnosed cardiac condition — making it the first cleared tool of its kind to run on routine ECGs already in widespread clinical use.

For healthcare practices, this clearance matters not because it changes how ECGs are acquired, but because it changes what can be done with the ECG data you already collect routinely.

For the full technical context on ECG-AI, see ECG-AI explained: what it changes.

Bottom line as of June 2026: Anumana's algorithm received FDA clearance on April 8, 2026, with prior Breakthrough Device Designation and selection for the FDA's Total Product Life Cycle Advisory Program pilot. It runs on a standard 12-lead ECG — no specialized equipment, no new data acquisition step. The practice workflow change is in what happens after the ECG is read: a new flag, a new referral trigger, and new documentation requirements.

Key Takeaways

• Anumana received FDA clearance for its ECG-AI cardiac amyloidosis algorithm on April 8, 2026 — the first and only clearance of this type for a standard 12-lead ECG, per Anumana's announcement.

• The algorithm received prior FDA Breakthrough Device Designation, indicating FDA recognized it as addressing an unmet need in a serious condition, per Anumana.

• Anumana's ECG-AI was among the first 15 devices in the FDA's Total Product Life Cycle Advisory Program (TAP) pilot, which provides enhanced FDA engagement through development, per Anumana.

• ECG-AI runs on a standard 12-lead ECG — 0 new devices required (Anumana).

• Cardiac amyloidosis is described as under-diagnosed and hard to spot — symptoms overlap with common cardiovascular conditions and definitive diagnosis requires specialized testing not included in routine workups — the algorithm's value is earlier detection at a point in the care pathway (the routine ECG) where the condition would previously go unrecognized, per Inside Precision Medicine.

• For cardiology practices and primary care practices ordering routine ECGs, the workflow change is at the output stage: the algorithm adds a risk flag to the ECG read that triggers a documentation and referral decision.

• Practices need a defined clinical protocol for acting on ECG-AI flags before deploying the tool — the algorithm generates a flag; the practice decides what to do with it.

Who Should Care

This post is for: practice administrators, clinical operations managers, and physician leads at cardiology practices, internal medicine groups, and primary care practices of 3-50 providers that order standard 12-lead ECGs as part of routine patient workup — and that have an EHR-integrated ECG workflow.

Current stack that makes this relevant: Any practice using a 12-lead ECG device with EHR integration (Epic, Athenahealth, eClinicalWorks, or similar), processing ECG orders through the EHR workflow, with a documentation and results-routing process for abnormal readings.

The pain this touches: Cardiac amyloidosis is frequently diagnosed late — often only after a patient has experienced significant functional decline. A routine ECG is already collected across a wide range of patient encounters. An FDA-cleared algorithm that adds a risk flag to that existing workflow provides earlier detection at no additional data-acquisition cost to the practice.

Red flags: If your practice does not order 12-lead ECGs routinely (some very small specialty practices), the algorithm does not apply to your immediate workflow. If your ECG device and EHR combination does not support third-party algorithm integration, implementation requires vendor coordination before any clinical benefit is accessible. If your practice has not established a clinical protocol for acting on AI-generated ECG flags, deploying the tool without one creates documentation and liability risk — the protocol is a prerequisite, not an afterthought. Practices in states with AI-in-clinical-decision-support disclosure requirements should verify disclosure language with their compliance team before go-live.

What the FDA Clearance Means (The Facts)

On April 8, 2026, Anumana announced FDA clearance of its ECG-AI algorithm for cardiac amyloidosis detection. The key documented facts:

Clearance type and scope: According to Anumana's announcement, the clearance is for software-as-a-medical-device (SaMD) that runs on a standard 12-lead ECG — the most common ECG format in clinical use. The algorithm is described as the first and only FDA-cleared tool of its type for this indication.

Breakthrough Device Designation: Anumana received FDA Breakthrough Device Designation prior to clearance, per Anumana's announcement. This designation is granted by FDA when a device addresses a serious or life-threatening condition and provides more effective diagnosis or treatment than currently available alternatives.

TAP Pilot selection: According to Anumana, the algorithm was among the first 15 devices selected for FDA's Total Product Life Cycle Advisory Program pilot — a program providing enhanced FDA engagement through the device's entire development and post-market lifecycle.

Clinical rationale: According to DI Cardiology, cardiac amyloidosis is 1 of the most consistently under-diagnosed cardiac conditions in clinical practice — its early symptoms overlap with more common cardiac conditions, and the algorithm enables detection at the routine 12-lead ECG without requiring any additional equipment or data acquisition step.

Anumana's ECG-AI algorithm is the first FDA-cleared cardiac amyloidosis detection tool running on a standard 12-lead ECG — with Breakthrough Device Designation and TAP pilot selection confirming FDA's recognition of unmet clinical need (Anumana).

According to Inside Precision Medicine, the FDA cleared the first — and as of June 2026, only — AI algorithm for cardiac amyloidosis detection, with 0 competing cleared products for this indication. The condition has historically been identified in only a small fraction of patients who have it, due to nonspecific symptoms and the specialized testing required for a definitive diagnosis. The algorithm's ability to run on a standard 12-lead ECG means it reaches patients at one of the most common diagnostic touchpoints in clinical practice without requiring any additional equipment investment.

ECG-AI Fast Facts for Practice Operators

What Changes in Practice Operations

The ECG workflow for most practices follows a defined sequence: provider orders ECG → technician or nurse acquires the tracing → ECG device transmits to EHR → cardiologist or provider reads and signs the result → result routes to the ordering provider and the patient record.

The ECG-AI algorithm inserts at the transmission step: the algorithm processes the tracing and adds a risk flag to the result before it reaches the reader. The reader sees both the standard ECG tracing and the algorithm's output. Nothing about data acquisition changes. The change is in what the reader is expected to do with an additional data point.

That sounds simple. In practice, it creates three new operational requirements:

1. Clinical Protocol for Acting on AI Flags

A positive or elevated-risk flag from the algorithm is not a diagnosis. It is a screening signal that requires a defined clinical response: acknowledge in the note, order confirmatory testing (cardiac MRI, technetium pyrophosphate scan, or referral to amyloidosis specialist), or document clinical reasoning for not pursuing further workup. Without a written protocol, each clinician handles the flag differently — creating inconsistency in documentation and potential liability exposure.

The protocol document is a clinical governance deliverable, not an IT deliverable. It needs sign-off from clinical leadership before the algorithm is deployed.

2. EHR Documentation and Results Routing

Once the protocol is defined, the EHR workflow needs to reflect it. In Epic, this means configuring a Best Practice Advisory (BPA) or a results routing rule that triggers when the ECG-AI flag reaches a defined threshold. In Athenahealth, the equivalent is a clinical decision support alert configuration. The documentation template for the ECG read needs a field for recording the AI flag value and the clinical response taken.

This is an EHR configuration task, not a clinical task — but it requires clinical input to define the threshold and response language before IT can build it.

3. Referral Coordination for Flagged Patients

Cardiac amyloidosis confirmation typically requires subspecialty evaluation (advanced heart failure, cardiac imaging, or hematology depending on amyloid type). A practice that begins flagging patients at volume needs a referral pathway — a defined subspecialty partner, a standard referral template, and a process for following up on referral completion and results.

Practices without an existing referral management workflow will need one before ECG-AI flags generate referral volume. This is the highest-friction operational gap for primary care and general cardiology practices that currently handle amyloidosis referrals sporadically.

Worked Example: Flagging and Referral Workflow at a Cardiology Group

Consider a 6-cardiologist group practice that reads approximately 400 ECGs per month across inpatient consults, outpatient clinic visits, and pre-procedure workups. As of June 2026, ECGs are acquired on a GE MAC 5500 HD system integrated with Epic, with results routed to the ordering provider's Epic inbox for sign-off.

After ECG-AI integration, the observation.created event in the Epic FHIR API fires when the algorithm processes each tracing. If the algorithm flags 2% of the 400 ECGs per month — that is 8 patients per month — the observation.created event triggers a BPA in Epic that surfaces the amyloidosis risk flag in the cardiologist's results inbox alongside the standard ECG read. For the 6-cardiologist group reading 400 ECGs per month, that is a manageable 8 flagged cases per month requiring a documented clinical response, but the protocol and EHR configuration must exist before the first flag fires or the workup decision defaults to the individual clinician with no guidance.

The US Tech Automations integration point is the referral coordination workflow: when the observation.created event fires in Epic with an amyloidosis flag above threshold, an automated task is generated in the referral coordinator's queue with the patient record, the flag value, and the standard referral template pre-populated. The coordinator reviews, calls the patient, and sends the referral — without manually identifying flagged ECGs from the inbox.

Implementation Checklist for Healthcare Practices

Estimated Workflow Time Impact (Illustrative)

These estimates are based on typical ECG workflow cycle times and the operational changes described above. Validate against your own practice data before making staffing decisions.

Sources: Illustrative estimates based on task type and described automation capabilities. Not sourced benchmarks.

Compliance and Billing Considerations

Source: AAPC — CPT 93000, CPT 93005, CPT 93010. Confirm specific codes with your coding team — this table is a reference starting point, not billing guidance.

AI disclosure: Several states have enacted or are considering requirements for patient disclosure when AI-generated clinical decision support is used. As of June 2026, requirements vary by state. Practices should verify applicable state rules with their compliance team before go-live.

Documentation: Because the algorithm is FDA-cleared SaMD (not experimental), using it in clinical practice does not require an IRB protocol. However, the clinical reasoning for acting on or overriding the flag should be documented in the clinical note — both for clinical continuity and for liability protection.

Billing: As of June 2026, no dedicated CPT add-on code exists for AI-assisted ECG interpretation — the ECG read is billed under existing ECG interpretation codes (CPT 93000, 93005, or 93010, depending on which components the provider performs), per standard ECG billing guidance (AAPC). According to AAPC, these 3 ECG CPT codes — 93000, 93005, and 93010 — cover the full range of acquisition and interpretation billing combinations for a standard 12-lead ECG. If the AI flag triggers additional testing (cardiac MRI, nuclear pyrophosphate scan), those are billed under their existing codes. Practices should confirm billing guidance with their coding team and monitor for new CPT guidance as AI-in-clinical-decision-support evolves.

Medicare advantage and prior authorization: If flagged patients require confirmatory imaging, prior authorization workflows need to account for the new indication path. Practices with automated prior authorization workflows — including those running through US Tech Automations — should add cardiac amyloidosis workup as a trigger condition in the authorization routing logic.

Signal vs Speculation

Confirmed (sourced):

• Anumana FDA clearance for ECG-AI cardiac amyloidosis algorithm on April 8, 2026, per Anumana.

• Breakthrough Device Designation confirmed prior to April 8, 2026 clearance, per Anumana.

• TAP pilot selection: among first 15 devices in the FDA's TPLC program, per Anumana.

• Cardiac amyloidosis is frequently under-diagnosed due to nonspecific symptoms; algorithm enables detection at the routine 12-lead ECG, per Inside Precision Medicine and DI Cardiology.

Our read (12-36 month forecast):

If Anumana's commercial rollout reaches a meaningful fraction of the cardiology and primary care ECG workflow over the next 12-24 months, the referral volume to cardiac amyloidosis specialists will increase — which is the intended clinical outcome and also a planning input for subspecialty practices that will receive those referrals. Practices near academic medical centers with amyloidosis programs should establish formal referral agreements now.

The broader trend — FDA-cleared AI algorithms running on existing diagnostic data without new equipment — will expand beyond cardiac amyloidosis. ECG-AI for atrial fibrillation risk, left ventricular dysfunction, and other conditions are in various stages of development and regulatory review. The practice operations infrastructure built for the Anumana algorithm (EHR alert configuration, referral coordination workflow, documentation templates) is reusable for subsequent ECG-AI deployments. Building it once for amyloidosis is building it for the category.

The uncertain variable is payer coverage for confirmatory testing triggered by AI flags. If commercial insurers and CMS publish guidance on prior authorization for workups initiated by FDA-cleared ECG-AI flags, the referral coordination workflow becomes more predictable. As of June 2026, that guidance has not been published.

Frequently Asked Questions

What is ECG-AI and how does it differ from standard ECG interpretation?

ECG-AI is FDA-regulated software that analyzes 12-lead ECG tracing data using machine learning to detect conditions beyond standard rhythm analysis. Anumana's cleared algorithm specifically detects patterns associated with cardiac amyloidosis — a condition not typically identified on a routine ECG by standard clinical review. The algorithm outputs a risk flag; the clinician makes the diagnostic and treatment decision.

Does my practice need new equipment to use Anumana's algorithm?

No. The algorithm runs on data from a standard 12-lead ECG — the most common format already in use across cardiology and primary care. No new acquisition equipment is required. Integration requires that your ECG system can transmit data to the algorithm's processing layer, which depends on your ECG device manufacturer and EHR vendor.

What is cardiac amyloidosis and why is early detection important?

Cardiac amyloidosis is a condition where abnormal amyloid protein deposits accumulate in the heart muscle, causing progressive stiffening and heart failure. Inside Precision Medicine notes it is frequently under-diagnosed — partly because its early symptoms overlap with common cardiovascular conditions and partly because definitive diagnosis requires specialized testing not included in routine workups. Earlier detection enables earlier treatment, which is associated with better outcomes in the available literature.

How does the algorithm integrate with Epic or other EHRs?

Integration approaches vary by EHR vendor and ECG device manufacturer. The algorithm outputs a risk flag that must be captured as an observation or result in the EHR. In Epic, this typically involves a BPA (Best Practice Advisory) configured to fire when the flag exceeds a threshold. Your EHR vendor and the Anumana implementation team will define the specific integration method. For practices already running document and result routing workflows through US Tech Automations, the algorithm output can be treated as a triggering event in the same workflow fabric. See how to set up renewal reminders for medical practices for a related example of EHR-triggered workflow automation.

What happens if the algorithm flags a patient who does not have cardiac amyloidosis?

The algorithm generates a risk flag, not a diagnosis. A positive flag triggers a defined clinical response (per your practice's protocol), which may include confirmatory testing that rules out the condition. The clinician's note should document the flag, the clinical assessment, and the reasoning for the response taken. This is standard practice for any screening tool with imperfect specificity — the protocol design should account for the expected false-positive rate. For document collection workflow automation for flagged patients, see how to set up document collection for medical practices.

What should my practice do before implementing this algorithm?

Four prerequisites: (1) Confirm ECG device and EHR compatibility with Anumana's integration requirements. (2) Draft and approve a clinical protocol for acting on AI flags before any patient encounter uses the algorithm. (3) Configure EHR alerts and documentation templates to support the protocol. (4) Establish a referral pathway to a subspecialty partner for positive cases. Implementing without all four in place creates clinical and operational risk. See how to stop duplicate data entry in healthcare workflows for how to avoid documentation redundancy when adding new data fields.

Getting Operationally Ready

The ECG-AI clearance on April 8, 2026 introduces a new category of FDA-cleared diagnostic support at the routine ECG level. The practices that implement it well are those that treat it as a clinical operations project — not just an IT integration — with a defined protocol, configured EHR alerts, and an active referral pathway before go-live.

The referral coordination and documentation workflows that follow an AI flag are where operational efficiency matters most. Practices handling 200-500 ECGs per month will see meaningful referral coordination volume from even a small positive-flag rate. Having that coordination automated — flag fires, referral task generates, coordinator reviews and acts — is the difference between manageable throughput and inbox chaos.

For practices building or extending the EHR-to-workflow integration layer, the customer service AI agent workflows for healthcare covers patient communication, referral coordination, and appointment scheduling automation patterns that pair directly with ECG-AI flag workflows.

The algorithm is cleared. The clinical and operational infrastructure to act on its outputs is what your practice builds next.