Evaluation of Scopio Labs X100 Full Field PBS: The first high‐resolution full field viewing of peripheral blood specimens combined with artificial intelligence‐based morphological analysis

Abstract Background Current digital cell imaging systems perform peripheral blood smear (PBS) analysis in limited regions of the PBS and require the support of manual microscopy without achieving full digital microscopy. We report a multicenter study that validated the Scopio Labs X100 Full Field PBS, a novel digital imaging system that utilizes a full field view approach for cell recognition and classification, in a decision support system mode. Methods We analyzed 335 normal and 310 abnormal PBS from patients with various clinical conditions and compared the performance of Scopio's Full Field PBS as the test method, with manual PBS analysis as the reference method. Deming regression analysis was utilized for comparisons of WBC and platelet estimates. Measurements of WBC and platelet estimation accuracy along with the agreement on RBC morphology evaluation were performed. Reproducibility and repeatability (R&R) of the system were also evaluated. Results Scopio's Full Field PBS WBC accuracy was evaluated with an efficiency of 96.29%, sensitivity of 87.86%, and specificity of 97.62%. The agreement between the test and reference method for RBC morphology reached 99.77%, and the accuracy for platelet estimation resulted in an efficiency of 94.89%, sensitivity of 90.00%, and specificity of 96.28%, with successful R&R tests. The system enabled a comprehensive review of full field PBS as shown in representative samples. Conclusions Scopio's Full Field PBS showed a high degree of correlation of all tested parameters with manual microscopy. The novel full field view of specimens facilitates the long‐expected disengagement between the digital application and the manual microscope.


| INTRODUC TI ON
Complete blood count (CBC) with peripheral blood smear (PBS) is a rapid common test that serves as a screening tool offering insights into patients' clinical conditions and guiding further laboratory workup. Despite the good performance of CBC analyzers, their limited capacity to identify morphological variations and abnormalities of blood cellular components [1][2][3] led to the establishment of a set of rules to trigger manual blood smear review following the outcome of CBC tests that is specific for each hematology laboratory. 4,5 The need to improve and standardize white blood cell (WBC), red blood cell (RBC), and platelet recognition has led to the development of several digital cell imaging systems that utilize various algorithms and methods to automate PBS image analysis, including image segmentation, feature extraction and selection, and pattern classification. 6 To date, a single vendor predominates morphological digital analyzers in hematology laboratories worldwide. 6,7 While current digital image analyzers are constantly improving and expanding, the main drawback of this field is that only limited fields of view (FOV) from the PBS are available for review. 6,7 As a mitigation, many of the digitally analyzed samples are also manually reviewed under a microscope, specifically those that contain cellular abnormalities, as detailed in the ICSH recommendations. 6 Hence, the sometimes-redundant triangle of the PBS, the digital image analyzer, and the manual microscope cannot easily be breached utilizing current technologies.
Our multicenter study evaluated and validated the FDA cleared Scopio Labs X100 Full Field PBS system (Scopio's Full Field PBS), a novel digital PBS morphological analyzer with full field specimens view as described in Supplementary 1. We analyzed 645 peripheral blood specimens, of which 335 were normal CBC and 310 were abnormal CBC collected from patients with various infectious or neoplastic conditions and compared the WBC differential, RBC morphology evaluation, and platelet estimation performance by the Scopio's Full Field PBS, with traditional manual PBS analysis performed by experienced medical technologists according to the Clinical and Laboratory Standards Institute H20, 2nd addition (H20-A2). 8 Here, we report a high degree of correlation between the two methods among the WBC classes, RBC morphology evaluation, and platelet estimation, as well as repeatability and reproducibility. In addition, we include representative full field PBS scans for evaluation of the capacity of this novel approach.

| Scopio labs X100 system
Scopio Labs X100 system is manufactured by Scopio Labs, Tel Aviv, Israel. The system is operated by a browser-based application, namely, the application does not require specific software installed, and may be accessed from any workstation running a browser, inside the secure medical facility network or securely connected to it remotely. The system is based on a computational photography approach, where a series of low-resolution full field images of the specimen are acquired by low power/wide field objective, and reconstructed into a high-resolution full field image based on a physical model (Supplementary 1A). The system includes automated platelet location and pre-estimation, and WBC pre-classification by artificial intelligence (AI) based tools into the following three groups: 1) main WBC classes-neutrophils (including segmented and bands forms), lymphocytes, monocytes, eosinophils, basophils; 2) other WBC classes-immature myeloid cells (promyelocytes, metamyelocytes and myelocytes), blast cells, lymphocyte variant forms, plasma cells; 3) non-WBC classes: nucleated red blood cells and smudge cells.
Both WBC and platelet pre-classifications operate as a decision support system (DSS), requiring the operator to review the pre-classified data generated by the system, approve, or correct it. So far, DSS is the only mode cleared by the FDA for such analyzers. Specimens were collected and analyzed according to various conditions detailed in Table 1, resulting in a total sample size of 645 specimens. Six slides were prepared from each specimen.

| Patients and samples
The clinically abnormal specimens were collected from patients according to distinct clinical categories, as specified in the CLSI H20-A2. 8 The distribution of samples across sites is shown in Table 1, and the demographics of the patients sampled are summarized in Supplementary 2.

| Sample preparation
The method of sample preparation depended on site protocols and differed slightly for each site. At site #1, specimens were collected into a spray-coated K2 EDTA 3.6 mL vacuum tube (BD Vacutainer). PBS were prepared, within four hours and at room temperature, by the Sysmex SP-10 which is a fully automated hematology slide preparation and staining system, on glass slides

| PBS analysis area
Scopio's Full Field PBS locates an optimal analysis area for each PBS to include the monolayer area as well as the feathered edge (Supplementary 1D). On average, the monolayer part of the scan is 0.38 cm 2 , equivalent to 1000 high power fields (100X magnification).
The adaptive scan feature contributed to a robust morphological analysis of short and long smears. No restrictions were applied on the analysis area for the manual review.

| PBS evaluation
Scopio's Full Field PBS performs WBC analysis by an artificial intelligencebased classifier, in a decision support system (DSS) mode. A total of 645 specimens were analyzed for WBC differentials, RBC morphology and platelets estimation by two independent operators, at three sites, using Scopio's Full Field PBS (test arm) and manual microscopy (reference arm). All six operators that participated in the study were qualified and certified to perform PBS morphological analysis by their respective site requirements. In addition, each site had an arbitrator in case of disagreement between the two operators in the reference arm only.
200-WBC differentials were evaluated as described in the statistical analysis section. If less than 200 WBC were available for analysis in one PBS (eg, in the cases of leukopenia), additional slides from the same specimen were analyzed. As a DSS, Scopio's Full Field PBS results for the WBC differentials were approved (or modified where required) by the operators (Supplementary 1). In the reference arm, the operators performed a manual WBC differential using a manual microscope.
The platelet estimation was derived by automatically locating and counting platelets in 10 FOVs (Supplementary 1E), and multiplying the total count by a constant factor specific to each center and method. As a DSS, platelet detections were approved (or modified where required) by the operators. For the reference method, the operators manually counted platelets in 10 FOVs and calculated a platelet estimation.
Twenty-two parameters of RBC morphology were evaluated by the operators (Supplementary 4). In the test arm, the operators reviewed the digitally scanned PBS with an overlaid grid, with each grid cell proportionally representing a single high-powered manual microscope FOV. For the reference method, the operators manually evaluated the RBC morphology.

| Repeatability
Following CLSI's EP05-A3 Evaluation of Precision of Quantitative Measurement Procedures, 3rd Edition (CLSI's EP05-A3), 9 standardized "20 × 2 × 2" (20 days, 2 runs, 2 replicas) repeatability experiment was conducted for 15 test samples (8 normal, 7 abnormal) TA B L E 1 Peripheral blood smears' distribution and classification across three testing sites and representative samples which were randomly selected from within each clinical group. For each test sample, WBC pre-classification and platelet estimation results were analyzed with a two-way nested ANOVA, and standard deviation (SD) estimates with 95% confidence interval (CI) for the repeatability, between-run (within-day), between-day, and withinlaboratory variance components were calculated.

| Reproducibility
Following CLSI's EP05-A3, 9 a standardized "3 × 5 × 5" (3 devices, 5 days, 5 replicas) reproducibility experiment was conducted for 10 test samples (5 normal and 5 abnormal), which were randomly selected from within each clinical group. For each test sample, WBC pre-classification and platelet estimation results were analyzed with a two-way nested ANOVA, and SD estimates with 95% CI for the between-day (within-site), between-site, within-laboratory variance, and reproducibility components were calculated.  Figure 2A. Standard deviation (SD) and 95% confidence intervals (CI) estimates were constructed for each WBC subclass.

| Statistical analyses
Upper bound of the SD's 95% CI values for within-laboratory precision component were plotted for the different WBC types in the tested samples. All values were below the pre-defined acceptance criteria of 5% ( Figure 2B). For platelets, Figure 2C

| D ISCUSS I ON
The goal of our study was to assess and validate the Scopio Labs X100 Full Field PBS application. In addition to demonstrating a high degree of correlation with the manual method for WBC classification, RBC morphology evaluation, and platelet estimation, Scopio's Full Field breakthrough technology enables viewing of whole blood smears via a modern browser-based application, accompanied by pre-classification of WBC and platelets estimation (links in Table 1 and Figure 1). While the performance of existing digital microscopic systems in classification of WBC is generally adequate in a decision support system (DSS) mode, the option of full field specimen viewing during PBS analysis is an unmet need. 6,7 The correct RBC morphological classification is problematic, and the automated identification of highly informative RBC forms such as schistocytes or teardrop cells need to be significantly reviewed. 6 As noted by the ICSH recommendations, abnormal leukocytes may be under-represented in limited-area digital smear analyses of TA B L E 2 Comparison of distributional normal ranges (%) and morphological normal ranges (%) between the manual and digital methods across three sites  Previous automated digital microscopes fail to identify many of the specimens containing platelet clumps due to their narrow field of view, rendering manual microscope reviewing of samples with CBC-derived thrombocytopenia essential. 6,7,[13][14][15] Previous digital microscopy users report limitations with respect to both platelets and RBC analysis, 14 and manual PBS analysis is recommended in various conditions. 15

CO N FLI C T O F I NTE R E S T
Ben-Zion Katz is a consultant to Scopio Labs.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.