(357 days)
DM1200 is an automated system intended for in-vitro diagnostic use.
The body fluid application is intended for differential count of white blood cells. The system automatically locates and presents images of cells on cytocentrifuged body fluid preparations. The operator identifies and verifies the suggested classification of each cell according to type.
DM1200 is intended to be used by skilled operators, trained in the use of the device and in recognition of blood cells.
CellaVision DM1200 with the body fluid application automatically locates and presents images of nucleated cells on cytocentrifuged body fluid preparations. The system suggests a classification for each cell and the operator verifies the classification and has the opportunity to change the suggested classification of any cell.
The system preclassifies to the following WBC classes: Unidentified, Neutrophils, Eosinophils, Lymphocytes, Macrophages (including Monocytes) and Other. Cells preclassified as Basophils, Lymphoma cells, Atypical lymphocytes, Blasts and Tumor cells are automatically forwarded to the cell class Other.
Unidentified is a class for cells and objects which the system has pre-classified with a low confidence level.
Here's an analysis of the provided text, outlining the acceptance criteria and study details for the CellaVision DM1200 with the body fluid application:
Acceptance Criteria and Device Performance Study for CellaVision DM1200 with Body Fluid Application
The CellaVision DM1200 with body fluid application is an automated cell-locating device intended for in-vitro diagnostic use, specifically for the differential count of white blood cells in cytocentrifuged body fluid preparations. The system automatically locates and presents cell images, suggests a classification, and requires operator verification.
1. Table of Acceptance Criteria and Reported Device Performance
The provided document does not explicitly state pre-defined "acceptance criteria" as pass/fail thresholds for accuracy or precision. Instead, it presents the results of a method comparison study between the CellaVision DM1200 (Test Method) and its predicate device, CellaVision DM96 (Reference Method), for various leukocyte classifications. The implied acceptance criteria are that the DM1200 demonstrates comparable accuracy and precision to the predicate device.
| Cell Class | Acceptance Criteria (Implied: Comparable to Predicate) | Reported Device Performance (CellaVision DM1200 vs. DM96) |
|---|---|---|
| Accuracy (Regression Analysis: DM1200 = Slope * DM96 + Intercept) | ||
| Neutrophils | Regression where slope is close to 1 and intercept close to 0, with high R² | y = 0.9969x + 0.0050, R² = 0.9932 (95% CI Slope: 0.9868-1.0070, 95% CI Intercept: 0.0004-0.0096) |
| Lymphocytes | Regression where slope is close to 1 and intercept close to 0, with high R² | y = 0.9815x + 0.0016, R² = 0.9829 (95% CI Slope: 0.9656-0.9973, 95% CI Intercept: -0.0049-0.0081) |
| Eosinophils | Regression where slope is close to 1 and intercept close to 0, with high R² | y = 1.1048x - 0.0002, R² = 0.9629 (95% CI Slope: 1.0782-1.1314, 95% CI Intercept: -0.0007-0.0003) |
| Macrophages | Regression where slope is close to 1 and intercept close to 0, with high R² | y = 1.0067x - 0.0050, R² = 0.9823 (95% CI Slope: 0.9901-1.0232, 95% CI Intercept: -0.0125-0.0024) |
| Other cells | Regression where slope is close to 1 and intercept close to 0, with high R² | y = 0.9534 + 0.0032, R² = 0.9273 (95% CI Slope: 0.9207-0.9861, 95% CI Intercept: -0.0002-0.0065) |
| Precision/Reproducibility (Short-term Imprecision) | ||
| Neutrophils | SD % comparable between test and reference method | Test Method: Mean % 32.0, SD % 3.2; Reference Method: Mean % 31.6, SD % 3.4 |
| Lymphocytes | SD % comparable between test and reference method | Test Method: Mean % 30.1, SD % 5.6; Reference Method: Mean % 30.5, SD % 5.7 |
| Eosinophils | SD % comparable between test and reference method | Test Method: Mean % 0.6, SD % 0.7; Reference Method: Mean % 0.5, SD % 0.6 |
| Macrophages | SD % comparable between test and reference method | Test Method: Mean % 35.3, SD % 5.8; Reference Method: Mean % 35.5, SD % 6.2 |
| Other cells | SD % comparable between test and reference method | Test Method: Mean % 2.1, SD % 1.7; Reference Method: Mean % 1.9, SD % 2.5 |
The conclusion states that the short-term imprecision was found to be equivalent for the test method and the reference method, and the accuracy results (high R-squared values, slopes close to 1, and intercepts close to 0) demonstrate substantial equivalence.
2. Sample Size Used for the Test Set and Data Provenance
- Sample Size for Test Set: 260 samples.
- CSF: 62 samples
- Serous fluid: 151 samples
- Synovial fluid: 47 samples
- Data Provenance:
- Country of Origin: Not explicitly stated, but samples were collected from "two sites." Given the submitter is in Sweden, and the regulatory contact is in the USA, it's unclear if these sites were in Sweden, the USA, or elsewhere.
- Retrospective or Prospective: Not explicitly stated, but the description "collected from two sites" and then analyzed suggests a prospective collection or at least fresh samples for the study.
3. Number of Experts and Qualifications for Ground Truth
- Number of Experts: Not explicitly stated for the establishment of ground truth for the test set's initial classifications. However, the document mentions:
- "The results were then verified by skilled human operators." This indicates human review post-analysis by both the test and reference methods.
- The "Intended Use" section states: "DM1200 is intended to be used by skilled operators, trained in the use of the device and in recognition of blood cells."
- Qualifications of Experts: "Skilled human operators, trained in the use of the device and in recognition of blood cells." No specific professional qualifications (e.g., "radiologist with 10 years of experience") are provided.
4. Adjudication Method for the Test Set
The document states: "The results were then verified by skilled human operators." It does not specify a multi-reader adjudication method like 2+1 or 3+1. It implies a single operator verification for each result generated by both the test and reference methods.
5. Multi-Reader Multi-Case (MRMC) Comparative Effectiveness Study
No, a multi-reader multi-case (MRMC) comparative effectiveness study designed to measure the effect size of how much human readers improve with AI vs. without AI assistance was not explicitly described. This study was a method comparison between two devices (one with the new application, one being the predicate) with human verification. The device's function is to suggest classifications, which implies an assistive role, but the study design was not an MRMC study comparing human performance with and without AI.
6. Standalone (Algorithm Only) Performance Study
Yes, a standalone performance was implicitly done. The "Test Method" (CellaVision DM1200) "suggests a classification for each cell," meaning the algorithm performs an initial classification without human intervention. The reported accuracy metrics ($R^2$, slope, intercept) compare these suggested classifications to those obtained by the reference method (DM96, which also involves algorithmic preclassification). However, the study concludes with human verification of these results, suggesting the standalone performance without the "human-in-the-loop" step described in the intended use is not the final reported performance. The "Accuracy results" table (Table 3.3) and "Precision/Reproducibility" table (Table 3.4) reflect the device's performance before the final human verification step that might change classifications.
7. Type of Ground Truth Used
The ground truth used for the comparison was established by the predicate device (CellaVision DM96) with its own human verification, after undergoing "a 200-cell differential count... with both the test method and the reference method. The results were then verified by skilled human operators." Therefore, it's a form of expert-verified reference measurement. It is not pathology, or outcomes data.
8. Sample Size for the Training Set
The document does not explicitly state the sample size for the training set used to develop the CellaVision DM1200's classification algorithms. It mentions "deterministic artificial neural networks (ANN's) trained to distinguish between classes of white blood cells," but no details about the training data are provided within this summary.
9. How the Ground Truth for the Training Set Was Established
The document does not explicitly describe how the ground truth for the training set was established. It states that the ANNs were "trained to distinguish between classes of white blood cells," implying that a labeled dataset was used for training, but the process of creating these labels (e.g., expert consensus, manual review) is not detailed in this 510(k) summary.
§ 864.5260 Automated cell-locating device.
(a)
Identification. An automated cell-locating device is a device used to locate blood cells on a peripheral blood smear, allowing the operator to identify and classify each cell according to type. (Peripheral blood is blood circulating in one of the body's extremities, such as the arm.)(b)
Classification. Class II (performance standards).