Search Results

The Mazor X is indicated for precise positioning of surgical instruments or spinal implants during general spinal surgery. It may be used in open or minimally invasive or percutaneous procedures. Mazor X 3D imaging capabilities provide a processing and conversion of 2D fluoroscopic projections from standard C-arms into a volumetric 3D image. It is intended to be used whenever the clinician and/or patient benefits from generated 3D imaging of high contrast objects.

The Mazor X navigation tracks the position of instruments, during spinal surgery, in relation to the surgical anatomy and identifies this position on diagnostic or intraoperative images of a patient.

The Mazor X system combines robotic trajectory guidance with navigated surgical instruments (either guided or free hand navigation) to enable the surgeon to precisely position surgical instruments and/or implants according to predefined planning. With the imaging capabilities of the system, the user can also visualize the implants on the patients CT. Same as the predicate device the modified Mazor X consists of a workstation with dedicated software, the surgical system, navigation camera, accessories, instruments and disposable kits. The modified Mazor X, the subject of this 510(k) application, introduces software and hardware modifications to the Mazor X System cleared in 510(k) K230064.

This 510(k) clearance letter describes a device, the Mazor X System / Mazor X Stealth Edition, and its substantial equivalence to a predicate device. It includes detailed information about new AI-enabled features: "2D Automatic Measurements" and "Plan Assist." The following analysis focuses on the acceptance criteria and study details for these AI components as presented in the provided text.

Acceptance Criteria and Reported Device Performance

Device: Mazor X System / Mazor X Stealth Edition (Software enhancements for 2D Automatic Measurements and Plan Assist)

Detailed Study Information for AI-enabled Features

1. 2D Automatic Measurements (AI-enabled Feature)

• Sample Size for Test Set:
  • Clinical Evaluation: 146 AP images, 253 LAT images (for all spinopelvic parameters). Each specific spinopelvic parameter had its own sample size ranging from 24 to 126 images.
• Data Provenance for Test Set: Not explicitly stated (e.g., country of origin). The note about "U.S. certified spine surgeons and radiologists" for reference standard implies some U.S. involvement, but doesn't specify data origin. It is explicitly stated that these datasets were independent from the development dataset, with data source site-level separation to ensure data independence.
• Number of Experts for Ground Truth (Test Set): Three U.S. board-certified radiologists.
• Qualifications of Experts (Test Set): U.S. board-certified radiologists. (No years of experience provided).
• Adjudication Method (Test Set): Not explicitly stated, but implies individual evaluation by the three radiologists and comparison to AI measurements.
• MRMC Comparative Effectiveness Study: No, a standalone performance of the AI algorithm was evaluated against expert annotations.
• Standalone Performance: Yes, the accuracy of the endplate and femoral heads detection algorithms, and the accuracy of 13 spinopelvic parameters' measurements were evaluated.
• Type of Ground Truth (Test Set): Expert annotations/measurements. Each spinopelvic parameter's ground truth was comprised of annotations (endplate lines, endplate endpoints, and/or femoral head circles) performed by the three U.S. board-certified radiologists using the Mazor X device.
• Sample Size for Training Set:
  • Endplate Detection: 2327 AP images, 2651 LAT images.
  • Femoral Heads Detection: 2233 LAT images.
• How Ground Truth for Training Set was Established: Endplate line and Circle femoral heads annotations were performed by trained labelers within a quality-controlled environment. In case of multiple annotations, they were aggregated. More than 65% of the data was reviewed by U.S. certified spine surgeons and radiologists.

2. Plan Assist (AI-enabled Feature)

• Sample Size for Test Set: 326 screw plans from 25 spine images.
• Data Provenance for Test Set: Not explicitly stated (e.g., country of origin). It is explicitly stated that these datasets were separated at the site level from the development dataset to ensure data independence.
• Number of Experts for Ground Truth (Test Set): Three experts.
• Qualifications of Experts (Test Set): Not explicitly stated, but are referred to as "experts" who evaluated clinical acceptability.
• Adjudication Method (Test Set): Each AI-generated screw plan was evaluated by three experts for clinical acceptability.
• MRMC Comparative Effectiveness Study: No, a standalone performance of the AI algorithm was evaluated against expert judgment of clinical acceptability.
• Standalone Performance: Yes, the clinical acceptability of AI-generated screw positions was evaluated.
• Type of Ground Truth (Test Set): Expert judgment of clinical acceptability.
• Sample Size for Training Set: 160 studies (101 3D CT scans, 59 3D O-arm scans), 942 vertebrae, 5,795 unique screw placements, and over 23,000 screw plannings.
• How Ground Truth for Training Set was Established: Annotations for pedicle screw planning (referred to as "reference Screw planning") were manually performed by 34 qualified Surgical support technicians using a released Mazor X software.

Ask a specific question about this device