A Novel Automated Slide-Based Technology for Visualization, Counting, and Characterization of the Formed Elements of Blood: A Proof of Concept Study.
This article describes a potential new approach to the CBC in which a known volume of undiluted anticoagulated whole blood is precisely applied onto a microscope slide as a monolayer. The method uses the standard morphology of an air-dried blood film stained with a Romanowsky stain. The analysis is accomplished by multispectral digital imaging that allows precise simultaneous counting of all of the formed elements, while their digital images are analyzed for their morphology and classification. An operator can readily review the digitized images on a computer screen. Thus, the prototype instrument provides quantitative and morphologic information from the same sample of blood simultaneously. Compared with the existing combination of a flow cytometer, slide maker, and staining device with a stand-alone cell imager, the device is a single integrated system that eliminates electrochemical and laser components, reduces the number and volumes of reagents and liquid consumables supplies, and features a much smaller footprint in the laboratory. The time to complete a report that requires direct visual assessment of cellular morphology from a glass slide could be greatly reduced.
After the initial development work by the authors, (1) the new approach described here went into development to become a commercial instrument. Components of a prototype of that system are described here.
The CBC has evolved over time as new technologies have emerged. At present, more than 25 parameters appear in a typical CBC report, all of them derived from the quantitation and morphologic assessment of red blood cells (RBCs), WBCs, and platelets. Included are hematocrit and hemoglobin; the calculated RBC indices of mean cellular volume (MCV), mean corpuscular hemoglobin (MCH), and MCH concentration; and other derived values. Instrumentation has come a long way from the use of hemocytometer counting chambers for estimating cell numbers and the pulled, stained smear and microscope for evaluating normal and abnormal cells. From the introduction of single-channel Coulter Corporation particle counters in the 1950s to the current multichannel, multidetector flow cytometric systems, (2) accuracy has improved and costs have been reduced.
Todays flow-based systems are capable of performing 5or 7-part WBC differentials based on physical properties of blood cells, but they have no imaging component and still require addition specimen handling and microscopic "manual" review in many cases. (3) Because no images of cells are provided by a flow cytometric system, these systems require skilled operators with the ability to identify cell types and other parameters by interpreting their scattergram outputs. In terms of laboratory workflow, a significant problem with flow cytometry systems is that a large proportion of specimens give rise to "flags," indicating that further testing is required. "Flag rates" typically vary between 10% and 50%, depending on the flagging criteria policy established by the laboratory and the patient population it serves, with a median rate of 27%. (4) The most common reasons for "flagging" and retesting are the presence of abnormal cell types or abnormal cell morphology, or a too-low or too-high number of RBCs, WBCs, or platelets. In a recent survey by the College of American Pathologists, 35.7% of participants claimed to have learned additional information beyond what was available from the automated instrument's output by microscopic review of a peripheral smear. (4) A manual review is a labor-intensive and time-consuming step, requiring retrieval of the blood specimen tube, removal of blood from the tube, manual application of the sample onto a glass slide, and staining of the blood smear. The slide must then be examined under a microscope by a skilled hematology technologist or physician. Manual review delays the reporting time for the CBC, is labor intensive, and costs approximately 3 times the amount of the instrumental component of the CBC.
Some flow cytometry blood analyzer systems provide slide maker and stainer devices separately or in tandem with their automated CBC system. There are also stand-alone computer image analysis instruments that can be used for morphologic analysis after the slide is produced. Such additions make current systems more expensive; add to the complexity, size, and logistics of the CBC workstation; and require integrating the results of the imaging back into the initial CBC. Most significantly, this process means that the slide making, staining, and slide review are each done in sequence, lengthening the overall turnaround time to provide a result to the clinician.
Automated CBC instruments based on flow technology are considered to be mature, trustworthy, and efficient, and there have been no fundamentally different approaches introduced for several decades. Image analysis-based systems for the WBC differential count were first introduced by 4 major biomedical device companies in the 1970s. (5-8) They required separate preparation and staining of conventional blood smears, did not measure any other CBC parameters besides the WBC differential count, and had other limitations to their image analysis capabilities. (9) In addition, they were expensive. As a result, they did not survive competition from flow cytometry-based systems that came onto the market at that time; the flow cytometry-based systems were capable of doing some portions of the WBC differential, albeit without producing images of the cells. Recent versions of conventional smear-based image analysis systems have led to successful stand-alone instruments for the WBC differential, but they do not perform any of the quantitative measurements required for a CBC. (10)
NEW APPROACH TO INSTRUMENTATION FOR AN IMAGE-BASED CBC AND WBC DIFFERENTIAL
The prototype for a new, integrated system for the automated CBC and WBC differential described here overcomes certain intrinsic limitations and fixed operating requirements of flow cytometry. We describe below how the system works.
Racks of blood samples in standard ethylene diamine tetraacetic acid anticoagulated blood tubes are placed into the device for processing. For each tube in turn, a small volume (30 [micro]L) of blood is first withdrawn from the tube, after automated tube handling and mixing. From this volume, a blunt stainless steel needle withdraws an aliquot and transfers a precise volume (a nominal 1 [micro]L with 1% coefficient of variation) of blood directly onto a microscope slide (Figure 1). The blunt needle lays down adjacent rows of blood in a "monolayer" in which all cells can be distinguished and counted (Figure 2).
After the initial preparation, the slide is air-dried, fixed, and stained with a Romanowsky-type stain where the eosin solution and the methylene blue oxidized solution are applied sequentially, followed by rinsing. The stain formula was optimized to allow excellent visualization and assessment of the cell morphology from either the glass slide or from images on a computer screen, and to provide for consistent high-quality measurements by computerized analysis. An illustration of the staining mechanism used in the system is seen in Figure 3. The device stains the cells by infusing and then evacuating small amounts of the various reagents within a narrow space between the upside-down slide and a plastic surface.
After staining, the slide is transferred to the first of 2 digital imaging stations. At the first station, the digital imaging is performed at low magnification to count RBCs, WBCs, platelets, and nucleated RBCs. A very fast linear stage is used to move the slide, and images are acquired by a black-and-white charge-coupled device (CCD) camera while the specimen is illuminated with different wavelengths of light using 4 light-emitting diodes (LEDs). An image for each color is acquired sequentially and at different focus levels to allow dynamic tracking of focus while the slide is being analyzed. The optical systems are designed to provide high-quality images without the use of oil immersion. A view of the slide stage of the imaging assembly is shown in Figure 4.
After the low-magnification analysis is complete, the slide is transferred to the high-magnification imaging station. At this second imaging station at least 600 locations known to contain WBCs are digitized, again using LED illumination and a high-resolution, black-and-white charge-coupled device camera. From these images, a WBC differential of more than 600 cells is determined using image analysis techniques that take into consideration nuclear and cytoplasmic size, shape, texture, and optical absorption characteristics. For RBCs, cellular hemoglobin and volume are determined for approximately 20 000 RBCs, and potential RBC inclusions are noted. From the approximately 20 000 RBCs assessed, MCH is determined from optical absorption of specific wavelengths, and total hemoglobin is calculated from the MCH and total RBC count. The MCV is measured from the same RBCs by determining cellular volume using the size and optical absorption characteristics at multiple wavelengths of the individual RBCs. The MCV is determined from the average measurement of these cells, and hematocrit is calculated from the MCV and total RBC count. Distribution widths of the RBCs are calculated from the measured values of cellular volume. Additional analysis of the high-magnification images provides information on individual platelet volumes, allowing mean platelet volume to be determined.
The instrument is designed to move slides from station to station once a minute. The slide preparation and the 2 imaging stations each require 1 minute, whereas the staining mechanism requires 2 minutes of processing. For this reason, there are 2 stainers built in so that the entire system, with the 2 staining stations working in parallel on different patient samples, can provide results at the rate of 1 CBC with WBC differential per minute, for 60 patient samples per hour. In practice, the first sample in a batch requires 5 minutes for the preparation (1 minute), staining (2 minutes), low-magnification imaging (1 minute), and high-magnification imaging (1 minute). After that first sample, the remaining samples are processed and reported at a rate of 1 per minute.
A display is used to both monitor the operation of the instrument and to review data and images from "flagged" specimens requiring further review. This "viewing station" resides on a state-of-the-art computer with proprietary software.
The complete system is capable of measuring and reporting all standard CBC parameters and the WBC differential counts. The reportable data include a standard set of 23 parameters. Additionally, a reticulocyte count can be performed. When this is requested, an aliquot is mixed with a supravital stain before the slide is produced, stained, and imaged. Instrument messages or "flags" are similar to those found on flow systems, with the important distinction that when the operator "clicks" the mouse on a particular message on the computer screen, the program brings the operator immediately to the relevant images to allow review and resolution of the flag.
White blood cells, RBCs, and platelets are presented in image galleries (Figure 5). The WBCs are partitioned by cell type, including the 5 standard normal categories as well as a sixth category of unclassified cells (ie, cells not of the 5 normal types) requiring review by an operator. Individual cells can be magnified on the computer screen and reclassified by the operator as appropriate. Double-clicking on any WBC image zooms in on a particular cell in a field of neighboring cells, providing the operator an opportunity to evaluate abnormalities in a broader context. The WBC differential values are updated automatically if any WBCs are reclassified during this review.
Individual RBCs and platelets also may be shown isolated from their neighboring cells, which allows an entire gallery of those cells to be arranged and displayed. For example, RBCs can be sorted by size, shape, or hemoglobin content. In addition, the operator can sort platelets by volume.
This allows the operator to quickly recognize the presence of unusual RBC or platelet morphology.
Initial "proof of concept" studies were performed to compare the CBC results obtained on the prototype instrument with those from the Sysmex XE-5000 Hematology Analyzer (Sysmex Inc, Kobe, Japan). Specimens for the comparison studies were obtained from the hematology laboratory of a large teaching hospital (Tufts Medical Center, Boston, Massachusetts) and tested per protocols for discarded specimens that were approved by that institution's review board (IRB No. 7492). As part of the subsequent commercial development phase, the initial studies presented here are being extensively enlarged and all conventional measures validated for accuracy, precision, analytic and reporting ranges, interferences, and other characteristics. Documentation of those studies is beyond the scope of this initial presentation of the new technology.
For this study, specimens were analyzed from a wide array of deidentified adult and pediatric inpatients and outpatients with a range of medical conditions, including hematologic and other malignancies. Specimens were sought and collected for comparison that had very high or low results for various parameters, without regard to diagnosis.
The prototype instrument and the Sysmex XE-5000 Hematology Analyzer were first cocalibrated by being run for 2 consecutive days, processing 364 samples. During the 11 subsequent days, data were collected from a total of 1857 samples analyzed on both instruments, running first on the comparative analyzer, then on the new instrument. Both runs were performed within 4 hours of venipuncture. Of the 1857 samples thus analyzed, WBC differential data were available for 1263 samples. The Table shows the Pearson correlation coefficients for the major values of the CBC parameters and the percentages from the automated differential count. All parameters showed good to excellent correlation over a range of values spanning 1.79 X [10.sup.6]/[micro]L to 6.8 X [10.sup.6]/[micro]L; 0.01 X [10.sup.3]/[micro]L to 60.4 X [10.sup.3]/[micro]L; and 5 X [10.sup.3]/[micro]L to 922 X [10.sup.3]/[micro]L for RBCs, WBCs, and platelets, respectively. In Figure 6, scatter plots of the WBC, RBC, platelet, and neutrophil counts compare the new instrument to the comparative system. A high degree of correlation is noted in these plots. The full limits of the RBC, WBC, and platelet linearity range and reporting range are currently under investigation through extensive multicenter studies on the finalized version of the instrument.
A unique aspect of the new slide-based system is that its underlying methodology is based on cell identification and counting that are identical to those performed in conventional microscopy. The application of multispectral imaging and proprietary computer algorithms allows an accurate and precise assessment of all CBC and WBC differential parameters in a single instrument. In spite of the different methodology compared with flow-based instruments, all CBC and differential values appear to correlate well. Correlation coefficients and offsets between the new instrument and the comparative instrument show values very similar to those found when comparing 2 different flow-based analyzers. (11)
In the new image-based system, both the actual slide and digitized images are available to operators and clinicians for real-time or subsequent review, resolution of abnormalities, archiving, and staff education. The system's throughput of 60 CBCs with automated WBC differentials per hour, after an initial dwell time of 5 minutes, is comparable to current flow-based systems. It provides the opportunity for realtime, more efficient resolution of abnormalities, compared with the current practice of postanalysis slide making and staining, and either manual or automated WBC differential counting.
Our study has several limitations. We did not fully assess the effect of interfering substances on the CBC and automated differential count. These effects will be addressed fully in future reports employing the final commercial version of the product. However, it is intrinsic to our method that certain interferences that are important for flow-based CBC instruments be obviated with our image-based system. For example, regardless of their nature, it is important that substances that can lead to clumping of formed elements that might affect flow cytometric measurements, such as cryoglobulins or other RBC agglutinins, elevated fibrinogen, macroglobulins, or naturally occurring factors that lead to platelet clumping, rouleaux, or other conditions such as polycythemia or thrombocytopenia, be recognized and corrected for by image analysis. Rouleaux formation, for example, can be observed directly. As mentioned previously, RBCs in contiguity or overlapping are detected and counted. Similarly, individual platelets within small clumps can be counted directly by the instrument software. The MCH, MCV, and inclusions are determined using single RBCs found in the more than 600 high-resolution images. The generally recognized RBC inclusion bodies can be identified by the image recognition software, with accuracy dependent on their prevalence and stain quality.
Further, precision studies have been performed on a limited number of specimens. Results conformed to conventional standards (data not shown), with coefficients of variation similar to those of flow cytometry-based systems. Large-scale multicenter studies designed to establish definitive precision as well as accuracy, linearity, and reportable ranges for each CBC parameter using the fully developed system are underway. Comprehensive evaluation of recognition and classification of various RBC shape abnormalities and how to quantify them are part of ongoing clinical studies.
To our knowledge, we describe here, for the first time, a novel method for a monolayer preparation of a known volume of peripheral blood onto a microscope slide and an image analysis-based approach for determining the CBC and WBC differential counts from the slide.
Please Note: Illustration(s) are not available due to copyright restrictions.
(1.) Winkelman J, Tanasijevic M, Zahniser D, inventors; Roche Diagnostics Hematology Inc, assignee. Method for determining a complete blood count on a white blood cell differential count. US patent 8815537. August 26, 2014.
(2.) Hematology analyzers product guide. CAP Today. December 2014:21-36.
(3.) Aller R. High volume hematology analyzers. CAP Today. December 2005: 34-50.
(4.) Novis DA. Laboratory productivity and the rate of manual peripheral blood smear review. Arch Pathol Lab Med. 2006; 130(5):596-601.
(5.) Miller MN. Design and clinical results of Hematrak: an automated differential counter. IEEE Trans Biomed Eng. 1976; 23(5):400-405.
(6.) Cotter DA, Sage BA. Performance of the LARC Classifier in clinical laboratories. J Histochem Cytochem. 1976; 24(1):202-210.
(7.) Daoust PR. The clinical detection of variations in the concentrations of normal leukocyte types: a laboratory comparison of 100-cell manual differential counts on wedge smears and 500-cell counts by the ADC-500. Blood Cells. 1980; 6(3):489-496.
(8.) Marchand A, VanLente F, Galen RS. Automated differential leukocyte counters: a comparison of three systems. J Clin Lab Autom. 1983; 3:19-26.
(9.) Winkelman JW, Morris MN, O'Leary M. Spuriously elevated band counts with an automated differential counter. J Clin Lab Autom. 1963; 3(6):401-404.
(10.) Cornet E, Perol JP, Troussard X. Performance evaluation and relevance of the CellaVision DM96 system in routine analysis and in patients with malignant hematological diseases. Int J Lab Hematol. 2008; 30(6):536-542.
(11.) Meintker L, Ringwald J, Rauh M, Krause SW. Comparison of automated differential blood cell counts from Abbott Sapphire, Siemens Advia 120, Beckman Coulter DxH 800, and Sysmex XE-2100 in normal and pathologic samples. Am J Clin Pathol. 2013; 139(5):641-650.
James W. Winkelman, MD; Milenko J. Tanasijevic, MD, MBA; David J. Zahniser, PhD
Accepted for publication March 15, 2017.
Published as an Early Online Release April 19, 2017.
From the Department of Pathology, Brigham and Women's Hospital, and Harvard Medical School, Boston, Massachusetts (Drs Winkelman and Tanasijevic); and Roche Diagnostics Hematology, Boston (Dr Zahniser).
Drs Winkelman, Tanasijevic, and Zahniser were founders of Cell Imaging LLC. The company received National Institutes of Health-Small Business Innovation Research funding. Its technology was licensed to Constitution Medical Inc, which was subsequently acquired by Roche Holdings. Drs Winkelman and Tanasijevic received payments from Roche Holdings for the patent rights, development milestones, and consulting. A royalty agreement is in place. Dr Zahniser received payments from Roche Holdings for the patent rights and development milestones. A royalty agreement is in place. Dr Zahniser became an employee of Constitution Medical Inc and Roche Diagnostics Hematology Inc, and received, in addition to the above, salary and stock options.
Reprints: Milenko Tanasijevic, MD, MBA, Brigham and Women's Hospital, 75 Francis St, Amory Bldg 215 A, Boston, MA 02115 (email: firstname.lastname@example.org).
Caption: Figure 1. A microscope slide is moved under a blunt needle, providing a constant flow of whole blood. Typically, 1 [micro]L of blood is deposited.
Caption: Figure 2. A, The aliquot of whole blood is applied onto the microscope slide in adjacent rows, with all cells analyzable by computer imaging algorithms. B, Magnification of a section of the slide (original magnification X10).
Caption: Figure 3. Rapid staining is enabled by using separate blue and red components, while applying small volumes of reagents.
Caption: Figure 4. A high-speed servo motor stage moves the microscope slide under the imaging optics.
Caption: Figure 5. For white blood cell, red blood cell, and platelet review, individual cells are shown as a gallery of cells. A cell can be enlarged and displayed with its surrounding cells, as shown.
Caption: Figure 6. Comparability for the major parameters of (A) white blood cell (WBC), (B) red blood cell (RBC), (C) platelet (PLT), and (D) % neutrophil (NEUT) counts. The y-axis shows the new instrument's results; the x-axis shows the comparative flow-based system.
Correlation of the Major Complete Blood Count (CBC) and Differential Parametersa Parameter Pearson r Value White blood cells 0.99 Red blood cells 0.99 Hemoglobin 0.99 Hematocrit 0.99 Mean cellular volume 0.90 Mean corpuscular hemoglobin 0.97 Platelets 0.98 Mean platelet volume 0.87 % Neutrophils 0.98 % Lymphocytes 0.97 % Monocytes 0.76 % Eosinophils 0.96 % Basophils 0.63 (a) A total of 1857 samples were included in the CBC correlation, and 1263 in the white blood cell differential correlation.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Original Article|
|Author:||Winkelman, James W.; Tanasijevic, Milenko J.; Zahniser, David J.|
|Publication:||Archives of Pathology & Laboratory Medicine|
|Date:||Aug 1, 2017|
|Previous Article:||Cutaneous and Superficial Soft Tissue [CD34.sup.+] Spindle Cell Proliferation.|
|Next Article:||Best Practices in Immunohistochemistry in Surgical Pathology and Cytopathology.|