AI system
- System
-
HEART-SAFE photoplethysmography cardiac-arrest detection algorithm applied to Samsung Watch5 Pro sensor data
- Model
-
HEART-SAFE, a rule-based algorithm that compares pulse amplitude with thresholds.
- Release
-
Exact release not established.
- Application developer
-
HEART-SAFE project (Amsterdam UMC with Utek B.V. and MiBida) for the algorithm; Samsung for the Watch5 Pro
- Role
-
Not assessed
- Configuration
-
Researchers applied HEART-SAFE after the event to data from a Samsung Watch5 Pro. The algorithm did not run live or send an alert.
- Inputs and outputs
-
Not assessed
- Autonomy
-
Not assessed
- Workflow
-
A consumer smartwatch collecting green-light PPG through the project's own data-collection software. Reference ECG, arterial pressure and clinical PPG to time the arrest. Retrospective application of the HEART-SAFE peak-detection algorithm with minimum-amplitude safeguards.
- Human oversight
-
Not assessed
- Relation to AI
-
The algorithm was applied after the fact to exported green-PPG data. It marked the arrest when residual pulse amplitude fell below its thresholds, shortly before the monitor flatlined, with no false alarms. It was not running on the watch and triggered nothing. Only the developers' own papers describe the event. They describe peak detection with amplitude safeguards rather than machine learning. Samsung's role is limited to the hardware.
- AI contribution
-
The algorithm was applied after the fact to exported green-PPG data. It marked the arrest when residual pulse amplitude fell below its thresholds, shortly before the monitor flatlined, with no false alarms. It was not running on the watch and triggered nothing. Only the developers' own papers describe the event. They describe peak detection with amplitude safeguards rather than machine learning. Samsung's role is limited to the hardware.
Benefits
- Outcome identifiers
-
Not assessed
- Benefit
-
Amsterdam UMC case: a Samsung Watch5 Pro worn after treatment withdrawal captured PPG; the HEART-SAFE algorithm, run afterwards, marked circulatory arrest in step with the ECG; the patient died.
- Category
-
Improved Task Accuracy
- Status
-
Not assessed
- Beneficiary groups
-
People. Society.
- Beneficiaries
-
Not assessed
- Reach
-
Not assessed
- Distribution and access
-
Not assessed
- Duration
-
Not assessed
- Expected?
-
Yes
- Costs and harms
-
Not assessed
Measurement
- Studies
-
Not assessed
- Study design
-
Not assessed
- Measurement
-
Not assessed
- Setting and sample
-
Not assessed
- Measures
-
Not assessed
- Comparison
-
the clinical reference signals (ECG, invasive arterial pressure and clinical photoplethysmography) recorded alongside the watch
- Results
-
Result: The algorithm accurately identified the moment of cardiac arrest in concordance with the clinical reference signals.
- Additional value from AI
-
Not assessed
- Limitations
-
Not assessed
- Denominators
-
Not assessed
- Materials for replication
-
Not assessed
Money
- Kind of value
-
Not assessed
- Amount
-
Not assessed
- Currency and prices
-
Not assessed
- Period and unit
-
Not assessed
- Realization
-
Not assessed
- Recipients and payers
-
Not assessed
- Gross or net
-
Not assessed
- Included costs
-
Not assessed
- Comparison
-
Not assessed
- Method
-
Not assessed
- Return and payback
-
Not assessed
People and places
- Participants
-
Developer: HEART-SAFE project (Amsterdam UMC with Utek B.V. and MiBida) for the algorithm; Samsung for the Watch5 Pro. Operator: Amsterdam UMC intensive care unit (feasibility evaluation with institutional review board approval).
- Jurisdictions
-
Benefit: Amsterdam, the
Netherlands. Developer: Netherlands. Operator: Netherlands. - Ownership and control
-
Not assessed
- Dependencies
-
Not assessed
- Cross-border connections
-
Not assessed
- Geopolitical implications
-
Not assessed
Follow-up
- Conditions for replication
-
A live on-device implementation, which the authors deliberately did not use. Specificity high enough not to overburden emergency services, which the 2025 study says is still needed. Transfer from a slow, bradycardic end-of-life arrest to sudden arrhythmic out-of-hospital arrests, which the authors say the case does not represent.
- Replication attempts
-
1 follow-up entries. Benefit recurred: 1.
- Later findings
-
Nothing dated after the January 2026 case report was found; the group's proof-of-principle study predates it and the accompanying editorial reports no trial.
- Availability
-
Not assessed
- Institutional decision
-
Not assessed
- Grounds
-
Not assessed
- Action and result
-
Not assessed
- Success criterion
-
Not assessed
Evidence
- Source numbers
-
AIBD-0299-source-0001–0005
- Citations
-
Numbered references below.
- Language
-
Not assessed
- Source relationships
-
Developer, Independent
- Original evidence
-
Not assessed
- Access and versions
-
Not assessed
- Supported findings
-
Not assessed
- Evidence links
-
Not assessed
- Evidence strength
-
Not assessed
- Missing evidence
-
Not assessed
Grades
- Assessed benefits
-
People. Society.
- Depth
-
Not assessed
- Reasons
-
People: The patient died as expected under palliative withdrawal. The detection was retrospective and no alert, treatment or resuscitation followed. Society: The authors present the case as a step toward real-world detection of unwitnessed arrests. No deployment or survival outcome exists.
- Observed or intended
-
Not assessed
- Rubric
-
Not assessed
- Assessor and date
-
Not assessed
- Alternative assessments
-
Not assessed
- Fill
-
The makers' claim
Entry details
- Kind
-
Not assessed
- Related entries
-
Not assessed
- Field
-
Emergency medicine / resuscitation. Research use of a consumer smartwatch for automated cardiac-arrest detection.
- Purpose
-
The HEART-SAFE project builds device-independent software to detect circulatory arrest from smartwatch photoplethysmography and alert emergency services. This case tests the detection half on a true arrest.
- Dates
-
Date basis unassessed: 2026 (precision unassessed). Updated: 21 October 2026.
- Stage
-
Not assessed
- Version
-
SHA-256: 635582fad217…
References
- Other identifiers
-
Not assessed
- Category mappings
-
Not assessed
- Brief
-
Not assessed
- Revisions
-
5 revisions. Latest: 21 October 2026.
Sources
-
Peer-reviewed study Developer
van den Beuken WMF, Nideröst B, Goossen SA, Kooy TA, Demirtas D, Autar D, Loer SA, Eberl S, van Halm VP, Winkler BE, van Schuppen H, Tuinman PR, Schwarte LA, Schober P (2025). Automated cardiac arrest detection and emergency service alerting using device-independent smartwatch technology: proof-of-principle. Resuscitation 213:110657; doi:10.1016/j.resuscitation.2025.110657.
Across all three groups, 28 of 31 circulatory arrests were correctly identified, sensitivity 90.3% (95% CI: 74.2%-98.0%), and hour-level specificity was 94.1% (95% CI: 71.3%-99.9%).
-
Case report Developer
van den Beuken WMF, Tuinman PR, Nideröst B, Goossen SA, van Schuppen H, Loer SA, Schwarte LA, Schober P (2026). First automated detection of a cardiac arrest using a commercially available smartwatch: a case report. Resuscitation Plus 28:101247 (PMC12914423). Authors are the algorithm's developers (HEART-SAFE).
The algorithm accurately identified the moment of cardiac arrest in concordance with the clinical reference signals.
a commercial consumer-grade smartwatch, Samsung Watch5 Pro, was placed on the patient's right wrist approximately 40 min after WLST.
A key limitation of this report is that the algorithm was applied retrospectively rather than in real time.
-
Other Independent author
Veigl C, Billig S (2026). A digital witness for cardiac arrest: are wearables ready for real-world use? Resuscitation Plus 28:101246 (editorial accompanying the case report; PMC12914390).