| K-number | K251096 |
| Device name | PeekMed web |
| Applicant | Peek Health, S.A. |
| Product code | QIH |
| Device class | Class II |
| Decision date | Jul 14, 2025 |
| Decision | Substantially Equivalent |
| Regulation | 892.2050 |
PeekMed web is a cloud-based medical software system designed to help surgeons perform pre-operative planning for orthopedic procedures such as hip, knee, upper limb, and foot surgeries. The system imports patient medical imaging studies (X-rays, CT scans, and MRI), displays them in 2D or 3D environments, performs measurements and overlays digital representations of surgical implants, and generates final planning reports. It is intended for use by qualified healthcare professionals in planning surgical procedures on adult patients with diagnosed injuries or disabilities.
The subject device is substantially identical to its predicate (K250042) in software architecture, workflow, and core functionality. Both are cloud-based systems requiring internet connection, supporting automatic or manual pre-surgical planning with 2D/3D visualization, model positioning, dimensioning, and digital overlap of prosthetic templates. The key technical difference is that the subject device includes a new machine learning model variant for knee segmentation that additionally supports MRI imaging, whereas the predicate supported only X-ray and CT scans for knee segmentation.
Not stated in this summary.
The device is substantially equivalent because it shares identical indications for use, intended user population (adult patients, healthcare professionals), anatomical regions (hip, knee, upper limb, foot), contraindications (none), and core functional features with its predicate. Both devices perform the same workflows, have the same software architecture, and generate comparable planning outputs. The addition of MRI support for knee segmentation via a new ML model variant does not constitute an intended-use change because the development, verification, validation, and deployment processes are identical to the predicate, and all ML models met predefined acceptance criteria (segmentation DICE ≥90%, landmarking MRE ≤7mm, classification accuracy ≥90%, detection MAP ≥90%) through independent external validation on diverse datasets.
View the full FDA submission: accessdata.fda.gov