SpineAR SNAP (SyncAR Spine)

Authorization

K252054 documented held-out CT and MR segmentation performance for SPR.2.0.0 within an updated spine-planning and augmented-reality workflow.

Submission lineageK252054 → predicate K243623; terminated recall Z-0324-2026 concerned earlier software SPR.1.2.0, not the reviewed SPR.2.0.0 version.

Intended use

Provide preoperative spine planning on screen or in VR and intraoperative planning and visualization on screen or through a HoloLens2 AR display connected to validated navigation systems.

Input

CT and supported lumbar or lower-thoracic MR DICOM imaging; Third-party navigation position and orientation data; Surgeon-defined planning objects, virtual screws and guidance inputs

Output

Interactive three-dimensional spine scene; AI-generated editable vertebral and sacral segmentations; Virtual screw and decompression plans; Navigation and guidance cues on screen or in an AR headset

Use boundary

The virtual display must not be relied on as the sole source of absolute positional information and must be used with displayed two-dimensional stereotaxic information. The surgeon reviews and uses the segmentation and plan.

AI disclosure and function

AI/ML explicit

The summary explicitly identifies a 3D U-Net vertebral-segmentation architecture with iterative patch processing and a memory channel for sequential vertebrae.

The AI function segments vertebrae for visualization and surgical planning. The public summary does not identify continuous learning, a foundation model or autonomous navigation.

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

Validation

Separate retrospective performance evaluations for CT vertebrae, CT sacrum and MR vertebrae using held-out public or multi-site datasets, plus software, human-factors and cybersecurity testing.

Comparator

Expert-radiologist ground-truth segmentations and a prespecified lower 95% confidence bound for mean Dice above 0.8; K243623 served as the predicate.

Ground truth

Expert-radiologist segmentations in held-out Spine-Mets-CT-SEG, VerSe, TotalSegmentator and SPIDER subsets.

Representation

The CT set included 92 unique patients, 60% of scans from the US and 40% from Europe, balanced sex, ages 18-90 and three major scanner manufacturers. The MR set included 15 unique patients, both Philips and Siemens scanners and balanced male/female representation.

Reported performance

CT individual-vertebra mean Dice

0.912 (95% CI lower bound 0.907; 1,193 vertebrae)Retrospective segmentation overlap in 95 scans from 92 patients.

CT sacrum mean Dice

0.901 (95% CI lower bound 0.861; 38 scans)Held-out sacrum segmentation performance, separate from the primary vertebra set.

MR vertebra mean Dice

0.903 (95% CI lower bound 0.891; 208 vertebrae)Retrospective lumbar and lower-thoracic MR segmentation in 31 scans from 15 patients.

Limitations

• The public summary reports multiple overlapping validation units, so no single patient count is assigned.
• MRI validation included only 15 unique patients and was limited to lumbar and lower-thoracic regions.
• Segmentation performance does not establish navigation accuracy, screw placement, complications or patient outcomes.
• The AR display is explicitly not the sole source of absolute positional information.

It does not prove diagnostic performance, autonomous navigation, screw-placement accuracy, clinical superiority or improved patient outcomes.

PCCP / Change control

No authorized PCCP was identified for this submission in the reviewed FDA database record or decision material.

Absence of an identified PCCP does not prevent every future device change; it means no authorized plan was established in the reviewed public record.

Postmarket check

FDA Class II recall Z-0324-2026 concerned SpineAR/SyncAR SNAP software version SPR.1.2.0 and a possible secondary-image-layer misalignment under specific rotational-alignment conditions; the recall was terminated on 3 August 2026.

K252054 describes software version SPR.2.0.0. The earlier-version enforcement record is retained as a product-family signal and is not attributed as a recall of the reviewed version or submission.

Primary FDA sources

1. FDA AI-enabled device list — Submission K252054Confirms inclusion, list display name, company, lead panel, primary product code and decision date only.
2. FDA submission database — 510(k) K252054Confirms pathway, Orthopedic review panel, decision and No PCCP database flag.
3. FDA decision summary — 510(k) summary pp. 1-9; intended use pp. 1-3; AI model and datasets pp. 5-8Supports intended use, 3D U-Net description, held-out datasets, representation, performance results and position-display boundary.
4. FDA recall record — Recall Z-0324-2026Supports the prior SPR.1.2.0 product-family recall and termination date; does not attribute it to SPR.2.0.0 or K252054.

Sources checked Sep 9, 2026 · Next scheduled review Oct 9, 2026. “No exact-name result” is a bounded search finding, not proof that no safety event or recall exists.