(75 days)
The NIDEK Specular Microscope CEM-530 is a non-contact ophthalmic microscope, optical pachymeter, and camera intended for examination of the corneal endothelium and for measurement of the thickness of the cornea.
The NIDEK Specular Microscope CEM-530 which is the subject of this 510(k) is a modification to the NIDEK Specular Microscope CEM-530 cleared in K151706. The only change to the cleared device is to the software which has been revised to improve the accuracy of the automated analysis method. All other aspects of the cleared device remain unchanged. The NIDEK Specular Microscope CEM-530 provides non-contact. high magnification image capture of the endothelium enabling observation of the size and shape of cells. Information such as the corneal endothelial cell density(CD), the coefficient of variation of corneal endothelial cell area (CV), % hexagonality of cells (%HEX), is analyzed through the captured images. The captured images and analysis results of the endothelium are used to assist in intraocular or corneal surgery, postoperative follow-up, and corneal observation such as for endothelial disorders or the corneal state of patients who wear extended-wear contact lenses. Observation is possible in the central area (visual angle: 5°) and peripheral area (visual angle: 27°) using a periphery capture function as well as in the Center of the cornea. The captured images and analysis results can be printed on the built-in printer or optional video printer, or output to an external device over LAN connection. In addition to the specular microscopy, the corneal thickness can be optically measured in a non-contact method. The CEM-530 has auto-tracking and auto-shooting functions. Results can be printed using the the built-in thermal printer or captured images can be transferred to a filing system via LAN connection.
Here's a summary of the acceptance criteria and study details for the NIDEK Specular Microscope CEM-530, based on the provided text:
1. Table of Acceptance Criteria (Inferred from comparison with predicate) and Reported Device Performance
The acceptance criteria are implicitly defined by demonstrating substantial equivalence to the predicate device, Konan Medical, Inc. Cellchek Plus (K120264). The study aimed to show agreement and precision between the CEM-530's automated analysis and the Konan Cellchek Plus's manual center method. The tables provided present the direct comparisons that demonstrate this.
| Metric (Agreement Study) | Acceptance Criteria (Proxy: Performance of Konan Cellchek Plus) | Reported Device Performance (NIDEK CEM-530 Automated Analysis) - All Subjects (N=74) |
|---|---|---|
| Endothelial Cell Density (CD) | Mean (SD): 2738.6 (412.67) | Mean (SD): 2773.3 (343.10) |
| Mean Difference (CEM-530 vs. Konan) | Implicitly, a small difference and 95% LOA encompassing 0. | 34.7 (157.38) |
| 95% LOA (CEM-530 vs. Konan) | Implicitly, a small range. | (-280.1, 349.4) |
| Correlation (R) | Implicitly, high correlation. | 0.9296 |
| Coefficient of Variation of Endothelial Cell Area (CV) | Mean (SD): 29.6 (4.16) | Mean (SD): 27.4 (4.61) |
| Mean Difference (CEM-530 vs. Konan) | Implicitly, a small difference and 95% LOA encompassing 0. | -2.2 (4.03) |
| 95% LOA (CEM-530 vs. Konan) | Implicitly, a small range. | (-10.3, 5.9) |
| Correlation (R) | Implicitly, high correlation. | 0.5817 |
| % Hexagonality (%HEX) | Mean (SD): 62.6 (7.42) | Mean (SD): 68 (5.16) |
| Mean Difference (CEM-530 vs. Konan) | Implicitly, a small difference and 95% LOA encompassing 0. | 5.4 (8.52) |
| 95% LOA (CEM-530 vs. Konan) | Implicitly, a small range. | (-11.7, 22.4) |
| Correlation (R) | Implicitly, high correlation. | 0.1185 |
| Metric (Precision Study) | Acceptance Criteria (Proxy: Performance of Konan Cellchek Plus) | Reported Device Performance (NIDEK CEM-530 Automated Analysis) - All Subjects (N=45 for CEM-530, N=61 for Konan) |
|---|---|---|
| Endothelial Cell Density (CD) | ||
| Repeatability SD | 62.6 | 74.2 |
| Repeatability SD as % of Mean | 2.4% | 2.7% |
| Reproducibility SD | 95.5 | 83.7 |
| Reproducibility SD as % of Mean | 3.7% | 3.1% |
| Coefficient of Variation of Endothelial Cell Area (CV) | ||
| Repeatability SD | 2.7 | 1.7 |
| Repeatability SD as % of Mean | 8.4% | 6.2% |
| Reproducibility SD | 2.7 | 1.9 |
| Reproducibility SD as % of Mean | 8.5% | 6.8% |
| % Hexagonality (%HEX) | ||
| Repeatability SD | 5.3 | 3.7 |
| Repeatability SD as % of Mean | 8.7% | 5.5% |
| Reproducibility SD | 5.4 | 3.7 |
| Reproducibility SD as % of Mean | 8.9% | 5.5% |
2. Sample Size and Data Provenance
- Test Set (Effectiveness Population): 74 subjects
- Subgroups: 28 non-pathologic young eyes, 27 non-pathologic adult eyes, 19 pathologic adult eyes.
- Precision Population Subset: 45 subjects (15 from each subgroup).
- Data Provenance: Prospective clinical study conducted at one clinical site in the United States.
- Training Set: Not explicitly mentioned in this document for the new auto-cell count algorithm. However, the study states that "Endothelial image data captured on the CEM530(Ver1.09) in the previous study, CEM-530-US-0002 were imported for auto analysis based on a new auto-cell count algorithm." This implies that the algorithm was trained using data from the prior study.
3. Number of Experts and Qualifications for Ground Truth
The document does not explicitly state the number of experts used or their specific qualifications for establishing the ground truth of the test set against which the automated CEM-530 was compared.
However, for the agreement study, the CEM-530's automated analysis results were compared against "manual center method measurements performed with the Konan predicate device." This implies that the Konan device's manual measurements served as the comparative 'ground truth' for this specific comparison. It's not stated how many operators performed these manual measurements or their qualifications, but these are inherently human-derived and subject to human variability.
An "Additional Manual Comparison" was conducted by comparing the CEM-530 automated analysis to CEM-530 manual analysis methods across three machines/operators. This indicates that at least 3 operators were involved in generating the manual ground truth for this internal comparison. Their qualifications are not specified beyond being "operators."
4. Adjudication Method
The document does not specify an adjudication method like 2+1 or 3+1 for resolving discrepancies in ground truth establishment. Given that one of the ground truth comparators was the "manual center method measurements performed with the Konan predicate device" and the other was "CEM-530 manual analysis methods across three machines/operators," it suggests either:
- No formal adjudication process was used, and the direct manual measurements were considered the ground truth.
- If multiple operators performed the manual measurements on the Konan, their agreement would likely be part of the precision analysis but not explicitly an adjudication of a single measurement.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
- Was one done? Yes, in a sense. The study compares the NIDEK CEM-530's automated analysis against the Konan Cellchek Plus's manual method (predicate device) and also against the CEM-530's own manual analysis method. This involves multiple "readers" (automated algorithm vs. human operators) and multiple "cases" (subjects).
- Effect size of human readers improve with AI vs without AI assistance: The study focuses on demonstrating the equivalence of the automated CEM-530 to existing manual methods (Konan) and its own manual methods. It does not provide an effect size for how much human readers improve with AI assistance. Instead, it evaluates the standalone performance of the AI (automated analysis) in comparison to manual benchmarks. The precision ratios (e.g., Repeatability Ratio, Reproducibility Ratio) illustrate how the CEM-530's automated precision compares to the Konan's and its own manual precision, often showing better or comparable precision for the automated method for CV and %Hex, and somewhat higher (less precise) for CD in the CEM-530 auto vs. manual comparison.
6. Standalone Performance Study
Yes, a standalone performance study of the algorithm (automated analysis without human-in-the-loop performance) was done explicitly. The "Agreement study" and "Precision study" sections detail the performance of the NIDEK Specular Microscope CEM-530 using its automated analysis method. These results are then compared to:
- The performance of the predicate device, Konan CellChek Plus (manual center method).
- The performance of the NIDEK CEM-530's own manual analysis method.
7. Type of Ground Truth Used
The ground truth used was human-derived manual measurements / expert consensus. Specifically:
- For the comparison against the predicate, it was "manual center method measurements performed with the Konan predicate device."
- For the internal comparison, it was "CEM-530 manual analysis methods across three machines/operators."
8. Sample Size for the Training Set
The document does not provide the specific sample size for the training set. It mentions that "Endothelial image data captured on the CEM530(Ver1.09) in the previous study, CEM-530-US-0002 were imported for auto analysis based on a new auto-cell count algorithm." This indicates that a dataset from a prior study (CEM-530-US-0002) was used for training/development of the new algorithm, but the size of that dataset is not specified.
9. How the Ground Truth for the Training Set was Established
The method for establishing ground truth for the training set is not explicitly detailed in the provided text. However, given that the algorithm's purpose is to automate cell counts, it's highly probable that the ground truth for training data would have been established through meticulous manual cell counting and analysis by human experts, similar to how the comparison ground truth was established.
§ 886.1850 AC-powered slitlamp biomicroscope.
(a)
Identification. An AC-powered slitlamp biomicroscope is an AC-powered device that is a microscope intended for use in eye examination that projects into a patient's eye through a control diaphragm a thin, intense beam of light.(b)
Classification. Class II (special controls). The device, when it is intended only for the visual examination of the anterior segment of the eye, is classified as Group 1 per FDA-recognized consensus standard ANSI Z80.36, does not provide any quantitative output, and is not intended for screening or automated diagnostic indications, is exempt from the premarket notification procedures in subpart E of part 807 of this chapter subject to the limitations in § 886.9.