SafeOp 3: Neural Informatix System

Authorization

The submission expanded monitoring workflows and authorized defined future changes to SSEP classification and related functions under a PCCP.

Submission lineageK252842 → predicate K234092; PCCP authorized in K252842.

Intended use

Provide intraoperative neuromonitoring during surgical procedures by recording MEP, SSEP and EMG or assessing the neuromuscular junction.

Input

Evoked-potential and electromyography signals; Stimulation-site and intraoperative monitoring data

Output

MEP, SSEP, EMG and neuromuscular-junction monitoring information; SSEP baseline classification

Use boundary

The public summary supports monitoring functions and controlled classifier changes; it does not establish fewer neurological injuries or autonomous intraoperative decision-making.

AI disclosure and function

Algorithm / classifier

The reviewed summary describes an SSEP baseline-classification algorithm and stimulation-site-specific classifiers that may be retrained under an authorized PCCP.

Inclusion in FDA's AI-enabled device list and use of a classifier do not establish use of an LLM, foundation model or generative AI.

No foundation-model, LLM, or generative-AI function was identified in the reviewed FDA materials.

Validation

Software, electrical-safety, electromagnetic-compatibility and nonclinical performance verification supporting substantial equivalence; PCCP acceptance criteria govern future classifier changes.

Comparator

The K234092 predicate and prespecified performance criteria for controlled classifier modifications.

Ground truth

Defined software requirements and labeled intraoperative signal data used for classifier verification; no patient-outcome reference standard was reported.

Representation

Not reported as a clinical cohort because the clearance determination did not rely on clinical performance data.

Reported performance

PCCP SSEP classifier positive predictive value acceptance criterion

>=90%A threshold for validating a modified classifier, not a reported clinical-outcome effect.

PCCP SSEP classifier accuracy and F1 acceptance criteria

>=85% eachAcceptance criteria for controlled changes, not evidence of reduced injury.

Limitations

• FDA states that the substantial-equivalence determination was not based on an assessment of clinical performance data.
• Classifier acceptance criteria should not be reported as observed patient outcomes.
• The public summary does not validate autonomous surgical reasoning or intervention.

It does not prove fewer neurological injuries, superior outcomes, autonomous intraoperative judgment or foundation-model use.

PCCP / Change control

Authorized in K252842. Stimulation-site-specific SSEP baseline classification, SSEP state handling, tEMG workflow changes and ongoing SSEP classifier maintenance described in the clearance summary.

• Independent firewalled validation data
• Overall PPV >=90% and accuracy and F1 >=85% for the modified SSEP classifier
• Non-inferiority to the cleared classifier at 95% confidence
• Periodic retraining limited to approximately four updates per year

A PCCP authorizes bounded changes under predefined controls. It is not permission for unrestricted or continuously self-learning behavior.

Postmarket check

FDA Class II recall Z-2793-2026 concerns SafeOp 3 software v2.1 and three specified SafeOp Patient Interfaces; a software issue may interrupt triggered EMG and require reboot.

This is a component- and serial-specific recall record, not evidence that every SafeOp 3 device was recalled and not a patient-injury rate.

Primary FDA sources

1. FDA AI-enabled device list — Submission K252842Confirms inclusion, list display name, company, panel, product code and decision date only.
2. FDA submission database — 510(k) K252842Confirms pathway, decision and PCCP database field.
3. FDA decision summary — 510(k) summary pp. 1-7; PCCP section pp. 5-7Supports intended use, predicate, classifier description, PCCP controls and evidence boundary.
4. FDA recall record — Recall Z-2793-2026Supports the specified component recall; does not expand the affected scope beyond the official record.

Sources checked Sep 8, 2026 · Next scheduled review Oct 8, 2026. “No exact-name result” is a bounded search finding, not proof that no safety event or recall exists.