(391 days)
The GDx is a confocal polarimetric scanning laser ophthalmoscope that is intended for imaging and three-dimensional analysis of the fundus and retinal nerve fiber layer (RNFL) in vivo. The GDx and its GDx Variable Corneal Compensation (VCC) and GDx Enhanced Corneal Compensation (ECC) RNFL Normative Databases aid in the diagnosis and monitoring of diseases and disorders of the eye that may cause changes in the polarimetric retinal nerve fiber layer thickness. The GDx is to be used in patients who may have an optic neuropathy.
The GDx with ECC Retinal Nerve Fiber Layer Normative Database is a confocal scanning laser ophthalmoscope comprising an opto-mechanical scanning laser head unit and a computer. The device employs Scanning Laser Polarimetry (SLP) to measure the Retinal Nerve Fiber Layer (RNFL) thickness using polarized light.
The provided text does not explicitly state acceptance criteria in a quantitative format (e.g., target accuracy, sensitivity, or specificity). However, it describes the clinical evaluation conducted to demonstrate the device's substantial equivalence and improved performance compared to a previous algorithm.
Here's a breakdown of the information based on the provided text:
1. Table of Acceptance Criteria and Reported Device Performance
| Acceptance Criteria (Implied) | Reported Device Performance |
|---|---|
| Substantial equivalence to predicate devices for intended use. | The GDx with ECC Retinal Nerve Fiber Layer Normative Database is substantially equivalent to the predicate devices regarding intended use, operating principle, function, and materials. Clinical evaluation supports expanded indications for use and demonstrates substantial equivalence. |
| Ability to measure RNFL thickness. | The ECC and VCC algorithms correlated well, and their ability to measure RNFL thickness was similar. |
| Improved atypical scan rate compared to VCC algorithm. | The ECC algorithm generated a lower atypical scan rate than the VCC algorithm. |
| Improved compensation for anterior segment birefringence. | The ECC algorithm compensated for anterior segment birefringence better than the VCC algorithm. |
| Safety and effectiveness for intended use. | "All testing deemed necessary was conducted... to ensure that the device is safe and effective for its intended use..." (General statement) |
2. Sample Size Used for the Test Set and Data Provenance
The text doesn't explicitly define a "test set" in the context of typical AI model evaluation (i.e., a separate, unseen dataset for final performance evaluation). However, it mentions a model validation subset:
- Sample Size for Model Validation: One-third of the normal and glaucoma study eyes.
- Since 251 normal subjects and 215 glaucoma subjects were collected, the total number of subjects is 466.
- Therefore, the model validation subset would be approximately 466 / 3 ≈ 155 subjects.
- Data Provenance: The study was conducted at ten sites. The text does not specify the country of origin but implies a multi-center study likely within the US, given the FDA submission. The study collected data on prospective subjects for the normative database and glaucoma cohorts.
3. Number of Experts Used to Establish the Ground Truth for the Test Set and Qualifications of Those Experts
The text does not specify the number of experts used to establish ground truth for the model validation subset.
- For the normative database, subjects were "deemed representative of the normal population." This implies clinical assessment but no specific number of experts or their qualifications are provided.
- For the glaucoma subjects, they were identified as having "early, moderate and late stage glaucoma." This also implies clinical diagnosis, but the number and qualifications of experts are not stated.
4. Adjudication Method for the Test Set
The document does not describe an adjudication method (e.g., 2+1, 3+1) for establishing the ground truth or validating the model's performance. The selection of data for model building and validation was done "randomly... by statistical programs."
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
- No explicit MRMC comparative effectiveness study involving human readers with and without AI assistance is described. The study primarily compares two algorithms (VCC and ECC) against each other for measuring RNFL thickness. It does not evaluate the effect of AI assistance on human reader performance.
6. Standalone (Algorithm Only Without Human-in-the-Loop Performance) Study
- Yes, a standalone study was performed. The clinical evaluation focuses on the performance of the ECC algorithm itself, particularly its ability to measure RNFL thickness, its atypical scan rate, and its compensation for anterior segment birefringence, in comparison to the VCC algorithm. The "nerve fiber indicator (NFI)" was trained and validated, which represents the algorithm's output.
7. Type of Ground Truth Used
- Clinical Diagnosis/Assessment:
- For the normative database, subjects were "deemed representative of the normal population."
- For the glaucoma subjects, they were identified as having "early, moderate and late stage glaucoma." This indicates a clinical diagnosis based on the standard of care at the ten study sites. The specific criteria for diagnosis (e.g., visual field defects, optic disc appearance) are not detailed.
8. Sample Size for the Training Set
The "model building subset" was used for training:
- Two-thirds of the normal and glaucoma study eyes were randomly selected as the model building subset.
- Normal subjects: 251 * (2/3) ≈ 167 subjects
- Glaucoma subjects: 215 * (2/3) ≈ 143 subjects
- Total training set size: Approximately 167 + 143 = 310 subjects.
9. How the Ground Truth for the Training Set Was Established
- The ground truth for the training set (model building subset) was established in the same manner as described for the overall study:
- Subjects were identified as "normal" (deemed representative of the normal population) or as having various stages of "glaucoma" (early, moderate, and late stage). This implies clinical diagnosis based on established medical criteria at the ten study sites, but specific details on diagnostic methods and expert qualifications are not provided.
§ 886.1570 Ophthalmoscope.
(a)
Identification. An ophthalmoscope is an AC-powered or battery-powered device containing illumination and viewing optics intended to examine the media (cornea, aqueous, lens, and vitreous) and the retina of the eye.(b)
Classification. Class II (special controls). The device, when it is an AC-powered opthalmoscope, a battery-powered opthalmoscope, or a hand-held ophthalmoscope replacement battery, is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 886.9.