(111 days)
ARVIS® Shoulder is indicated for assisting the surgeon in the positioning and alignment of implants relative to reference alignment axes and landmarks in stereotactic orthopedic surgery. The system aids the surgeon in making intraoperative measurements and locating anatomical structures of the shoulder joint based on the patient's preoperative imaging. ARVIS® Shoulder is indicated for total shoulder arthroplasty using the Enovis AltiVate implant system.
ARVIS® Shoulder is a computer-controlled surgical navigation system intended to provide intra-operative measurements to the surgeon to aid in selection and positioning of orthopedic implant components. The subject device is the equivalent shoulder system of the predicate ARVIS® Surgical Navigation System used for indicated knee and hip arthroplasties. ARVIS® Shoulder combines software, electronic hardware and surgical instruments to intraoperatively track tools and locate anatomical structures based on the patient's preoperative imaging. The navigation platform uses the same electronic hardware, mounted on the surgeon's head and waist, as the predicate device. A new equivalent navigation application and a new equivalent surgical instrument set are provided to allow surgeons to navigate instruments in shoulder arthroplasty procedures. The ARVIS® Shoulder workflow involves CT based reconstruction of the patient's shoulder anatomy and preoperative planning to enable image-based navigation. The surgeon uses the plan data as guidance to navigate and help position shoulder instruments and implants. The preoperative planning platform uses Al-based automatic image segmentation and landmarking algorithms. The data used to train and test the algorithms was labeled and validated in advance by trained experts. The total data consisted of 300 CT scans (from 300 patients) acquired from candidates for shoulder joint replacement surgery. The cohort was partitioned into two disjoint subsets through random sampling, with 80% producing a training dataset and 20% constituting the test dataset. The training dataset consisted of 240 CT scans (from 240 patients). Patient ages ranged from 36 to 89 years (mean age of 70), with 46% male and 54% female. All CT scans were acquired using FDA cleared CT scanners. The navigation system is intended to be used with the Enovis AltiVate implant system. ARVIS® Shoulder displays measurements as described in Performance Claims.
Here's a breakdown of the acceptance criteria and the study that proves the device meets them, based on the provided text:
1. Table of Acceptance Criteria and Reported Device Performance
Device: ARVIS® Shoulder
Study Type: Validation of AI algorithms for automatic image segmentation and landmarking.
| Metric (Segmentation) | Acceptance Criteria (AC) | Reported Result |
|---|---|---|
| Scapula Avg DSC | > 0.9 | 0.981 |
| Scapula Avg MAD | ≤ 2mm | 0.229mm |
| Scapula Avg HD | ≤ 5mm | 0.824mm |
| Humerus Avg DSC | > 0.9 | 0.989 |
| Humerus Avg MAD | ≤ 2mm | 0.352mm |
| Humerus Avg HD | ≤ 5mm | 0.917mm |
| Metric (Landmarking) | Acceptance Criteria (AC) | Reported Result |
|---|---|---|
| Glenoid Center Mean ED | 1.79mm | |
| Glenoid Center SPCR | 95.0% | |
| Trigonum Mean ED | 1.86mm | |
| Trigonum SPCR | 95.0% | |
| Inferior Point Mean ED | ≤ 3.72mm | 2.11mm |
| Inferior Point SPCR | ≥ 75% | 94.9% |
| Medial Epicondyle Mean ED | 3.19mm | |
| Medial Epicondyle SPCR | 83.3% | |
| Lateral Epicondyle Mean ED | 3.29mm | |
| Lateral Epicondyle SPCR | 83.3% | |
| Neck Plane Position Mean ED | 2.01mm | |
| Neck Plane Position SPCR | 90.0% | |
| Neck Plane Orientation Mean AS | ≤ 10° | 8.70° |
| Neck Plane Orientation SACR | 86.7% |
2. Sample Size and Data Provenance for Test Set
- Sample Size: 60 CT scans (from 60 unique patients)
- Data Provenance: The CT scans were acquired from patients who were candidates for shoulder joint replacement surgery. The scans were acquired using FDA cleared CT scanners (Toshiba, Siemens, Philips, GE Medical Systems, Canon). The specific country of origin is not specified.
- Retrospective/Prospective: The text describes the data as having been used to train and test algorithms, and the cohort was partitioned into disjoint subsets. This suggests the data was retrospective (collected prior to the study for the purpose of algorithm development and validation).
3. Number of Experts and Qualifications for Ground Truth
- Number of Experts: Total of 3 experts.
- 1 trained engineer
- 2 orthopedic surgeons
- Qualifications:
- Trained Engineer: More than 2 years' experience with medical image processing.
- Orthopedic Surgeons: Subspecialty qualifications in upper limb surgery.
4. Adjudication Method for Test Set
The adjudication method described is: None (single review - approval).
The reference (ground-truth) label for each CT volume was obtained by a manual process, reviewed, and approved by the consensus of the trained engineer and the two orthopedic surgeons. This implies a single, agreed-upon ground truth rather than a process of resolving disagreements between multiple independent assessments.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
There is no mention of a Multi-Reader Multi-Case (MRMC) comparative effectiveness study being done to measure the effect of AI assistance on human readers. The validation focuses solely on the standalone performance of the AI algorithms against expert-established ground truth. Clinical testing was explicitly stated as "not required".
6. Standalone Performance Study
Yes, a standalone (algorithm only without human-in-the-loop performance) study was done.
The study compared the algorithm-generated outputs for segmentation (Dice Similarity Coefficient, Mean Absolute Distance, Hausdorff Distance) and landmarking (Euclidean Distance, Angular Separation, Successful Point and Angular Classification Rates) against manually established ground truth.
7. Type of Ground Truth Used
The ground truth used was expert consensus.
It was established through a manual process, reviewed, and approved by a trained engineer with medical image processing experience and two orthopedic surgeons with subspecialty qualifications in upper limb surgery.
8. Sample Size for Training Set
- Sample Size: 240 CT scans (from 240 unique patients)
- Total Data Pool: 300 CT scans (80% used for training, 20% for testing).
9. How the Ground Truth for the Training Set Was Established
The text states that "The data used to train and test the algorithms was labelled and validated in advance by trained experts." While it details the process for the test set's ground truth, it implies a similar method was used for the training set's ground truth by "trained experts", without providing specific numbers or identical qualification details as for the test set. Given the context, it's reasonable to infer a process of expert labeling, likely by similar qualified individuals, but the exact expert composition for the training set ground truth isn't explicitly detailed with the same specificity as for the test set.
§ 882.4560 Stereotaxic instrument.
(a)
Identification. A stereotaxic instrument is a device consisting of a rigid frame with a calibrated guide mechanism for precisely positioning probes or other devices within a patient's brain, spinal cord, or other part of the nervous system.(b)
Classification. Class II (performance standards).