(78 days)
Customize for ankle arthroplasty is intended to be used as a software interface to assist in:
- Visualization, modification, validation of the planning of total ankle arthroplasty
- Communication of treatment options
- Segmentation of CT-scan data
- 3D CAD models generation
- use x-ray scan information to enhance the planning of ankle arthroplasty
- Managing timeline and cases
Customize is not intended to be used for
- Spine surgeries
- Implant and instrument design
Experience in usage and clinical assessment are necessary for a proper use of the software. It is to be used for adult patients only and should not be used for diagnostic purposes.
Customize is intended to be used during the preparation of ankle arthroplasties. It visualizes surgical treatment options that were previously created based on 3D CAD files generated from multi-slice DICOM data from a CT scanner. It consists of a single software user interface where the physician can review the CAD files in 3D, CT and X-Ray images in 2D and modify the position and orientation of the different 3D objects. Customize includes an implant library with 3D digital representations of various implant models so that the right implant positioning and sizing can be achieved based on the physician's input. After approval by the physician, the treatment plan is saved on the server and can be used as a reference during surgery. It is also possible to export the treatment plan for further processing such as designing and manufacturing of patient specific devices (the design of those later is done using an external software).
Here's a breakdown of the acceptance criteria and the study details for the "Customize" device based on the provided FDA 510(k) summary:
Acceptance Criteria and Device Performance
The provided document describes performance testing for the "Customize" device, focusing on segmentation validation, repeatability and reproducibility (R&R) of segmentation, and 3D model generation accuracy. While explicit "acceptance criteria" in terms of specific numeric thresholds are not directly stated in a table format, the narrative implies that the device successfully met the requirements of these tests to demonstrate substantial equivalence.
Implied Acceptance Criteria and Reported Device Performance:
| Acceptance Criteria Category | Reported Device Performance |
|---|---|
| Image Segmentation | Successful segmentation of ankle anatomies using an AI-based algorithm. The AI produces a "temporary label" as an initialization for manual editing, implying accuracy sufficient for efficient human supervision. The study "demonstrat[ed] Customize has been validated for its intended use" for image segmentation. |
| Repeatability & Reproducibility (R&R) Study | Successful demonstration of repeatability and reproducibility in the segmentation of ankle anatomies. This indicates consistency in the device's output under similar conditions. The study "demonstrat[ed] Customize has been validated for its intended use" for R&R. |
| Accuracy of 3D Model Generation (Tibia/Talus) | Accurate generation of 3D models for the tibia and talus based on CT-scan data. The study "demonstrat[ed] Customize has been validated for its intended use" for 3D model generation accuracy. No specific measurement units (e.g., mm, %) are provided for accuracy, but the overall conclusion of substantial equivalence implies these outcomes were within acceptable limits for a medical image management and processing system. |
Study Details
2. Sample size used for the test set and the data provenance:
- Segmentation validation and R&R study: The document states the AI-based ankle algorithm was trained on a dataset of 126 medical images (CT scans). It doesn't explicitly state the size of a separate "test set" for validation, but for these types of studies, a portion of the overall dataset or a new, unseen dataset is typically used for testing. Without further information, it's difficult to distinguish a test set size beyond the training dataset described.
- Data Provenance: Not explicitly stated whether the data was retrospective or prospective, nor the country of origin. The mix of patients (30% eligible for ankle arthroplasty, 70% cadaveric cases) suggests a blend of clinical and research data.
3. Number of experts used to establish the ground truth for the test set and the qualifications of those experts:
- This information is not provided in the document. The general statement about "human supervision by a 3D-Side engineer" applies to the device's operational use (editing AI output), but not necessarily to the establishment of ground truth for the validation studies.
4. Adjudication method for the test set:
- This information is not provided in the document.
5. If a multi-reader multi-case (MRMC) comparative effectiveness study was done, If so, what was the effect size of how much human readers improve with AI vs without AI assistance:
- No, an MRMC comparative effectiveness study is not described. The document mentions that the AI produces an "initialization" that "requires manual editions" and human supervision, suggesting it functions as an assistive tool rather than a standalone diagnostic or decision-making system. The focus is on the software's performance metrics (segmentation, R&R, 3D model generation), not on comparing human performance with and without AI assistance.
6. If a standalone (i.e. algorithm only without human-in-the loop performance) was done:
- The description implies that standalone performance was assessed to some extent for the "initialization" step. The AI algorithm generates a "temporary label" which then undergoes "human supervision." The "performance testing included 1) segmentation validation" of the software, suggesting its initial output (before human intervention) was evaluated against a ground truth. However, the device's intended use is described as AI followed by human supervision, so the final, clinical performance relies on the human-in-the-loop.
7. The type of ground truth used (expert consensus, pathology, outcomes data, etc.):
- The document implies that the "corresponding segmentation" data was used as ground truth for training the AI. For the performance studies, it is logical that similar "ground truth" segmentations (likely created by experts, though their number and qualifications are not specified) would be used for comparison, especially for segmentation accuracy and 3D model generation. Pathology or outcomes data are not mentioned.
8. The sample size for the training set:
- 126 medical images (CT scans).
9. How the ground truth for the training set was established:
- The ground truth for the training set was established through "corresponding segmentation." This typically means that experts manually segmented the anatomical structures (bones in this case) on the CT scans, and these expert-generated segmentations were used as the gold standard to train the AI model. The details of who performed these segmentations (e.g., number of experts, their qualifications) are not provided.
§ 892.2050 Medical image management and processing system.
(a)
Identification. A medical image management and processing system is a device that provides one or more capabilities relating to the review and digital processing of medical images for the purposes of interpretation by a trained practitioner of disease detection, diagnosis, or patient management. The software components may provide advanced or complex image processing functions for image manipulation, enhancement, or quantification that are intended for use in the interpretation and analysis of medical images. Advanced image manipulation functions may include image segmentation, multimodality image registration, or 3D visualization. Complex quantitative functions may include semi-automated measurements or time-series measurements.(b)
Classification. Class II (special controls; voluntary standards—Digital Imaging and Communications in Medicine (DICOM) Std., Joint Photographic Experts Group (JPEG) Std., Society of Motion Picture and Television Engineers (SMPTE) Test Pattern).