Home Evaluation of computer-assisted digital microscopes for evaluation of blood films1)
Article Open Access

Evaluation of computer-assisted digital microscopes for evaluation of blood films1)

  • Carl Thomas Nebe EMAIL logo , Manfred Ossendorf , Gudrun Stamminger , Henk Jansen and Thorsten Zerfaß
Published/Copyright: September 4, 2013
Download Article (PDF)
Become an author with De Gruyter Brill
Submit Manuscript Author Information
Laboratoriumsmedizin
From the journal Laboratoriumsmedizin Volume 36 Issue 5

Abstract

Currently, automated microscopes are increasingly entering clinical and private labs, making necessary a critical review of their performance. Within two trials, the two most advanced systems were compared with a hematology analyzer, a manual whole blood differential and an immunophenotypic approach. The diverse handling of nuclear shadows leads to the largest differences, followed by different classification schemes for lymphocytes and judgement of red cells and platelets. The implementation of these instruments into hematological step by step diagnostic schemes will be discussed. Because of poor low count statistics, an automated hematology analyzer differential and an immunological differential were included for comparison.

Reviewed Publication:

Nebe C.T.

Keywords: automated microscopy; blood film; digital morphology; five-part differential

Introduction

For a few years, automated microscopes have been in development or in clinical use that are to differentiate panoptically stained blood smears by means of PC-based software. These systems are currently being used increasingly in the medical diagnostic routine operations of private laboratories and hospitals, where they are used primarily to reduce personnel costs. A number of other such laboratories are considering the purchase of such systems, since they are increasingly combined with hematology instruments and offered as a package.

Distribution is currently mostly done by equipment manufacturers of hematology systems, which additionally offer a closed coloring system that allows for automated smear and staining, albeit in a separate unit. The coloring can be done in the stainers or manually with extra care. The demands on the quality of the smears are high: The cell density in the smear must be extremely uniformly thin, the color very constant and without dye residues. If this is the case, the devices create scanned images of mostly leukocytes and erythrocytes, which are then automatically analyzed, classified and displayed to the user for review. The number of the photographed leukocytes is above the sum of the cells to differentiate (e.g., at about 140 when 100 is preset) predetermined by the user, so that also in case of artifacts in the smear a reserve exists for the exclusion of recorded artifacts (dye particles, etc.) or smudge cells, erythroblasts, macro platelets or other non-leukocytes. There are very few publications on these devices and therefore there is the question of how to meet the expectations and how the manual microscopic differential blood count can be replaced by an automated one. Currently there are two providers in the market; more systems are about to be introduced to the market. Such devices are integrated into a hematological staged diagnosis of blood counts for the purposes of a diagnostic screen, where conspicuous blood counts can be studied further by flow cytometric measurement of the five-part differential blood count in the traditional hematology unit (Figure 1). Targeted by developers is also the analysis of blood smears for malaria diagnosis, cytospin preparations and bone marrow cytology.

Figure 1 Staged diagnostics.
Figure 1

Staged diagnostics.

Materials and methods

In three studies at different times, the two systems DM1200 (manufactured by CellaVision, Lund, Sweden, distributed by Sysmex Deutschland GmbH, Norderstedt, Germany, Figure 2) and HemaCAM (developed by Fraunhofer Institute in Erlangen, Germany, distributed by Horn-Imaging, Aalen, Germany, Figure 3) were examined (CellaVision in Chemnitz and Frankfurt, HemaCAM in Frankfurt only). The results were compared with the manual differentiation as well as the cytometric differential blood count of hematology instruments (CellDyn 3500, Abbott, Santa Clara, CA, USA) and the immunological blood count after flow cytometric immunophenotyping (FC500, Beckman Coulter, Miami, FL, USA).

Figure 2 Setup of the DM1200 from CellaVision.
Figure 2

Setup of the DM1200 from CellaVision.

Figure 3 HemaCAM system from Fraunhofer, distributed by Horn-Imaging.
Figure 3

HemaCAM system from Fraunhofer, distributed by Horn-Imaging.

Included were 300 (DM1200 study) and/or 326 (HemaCAM study) slides of unselected, consecutive samples of the specialized hematological laboratory at the Onkologikum in Frankfurt, Germany. Excluded were few aged blood samples with morphological changes, which do not allow for proper differentiation, and acute leukemias.

The anticoagulant of choice for hematology still is EDTA, which ensures an effective inhibition of platelet aggregation, but also leads to more rapid aging of the neutrophilic granulocytes, which can occur after only 6–8 h – this may differ from person to person and can be influenced significantly by environmental conditions (temperature, vibration, etc.). Heparin blood produces purple haze and more or less clear platelet aggregates or blood platelets attaching to neutrophils, and therefore is, despite a slightly better vitality of granulocytes, generally unsuitable. No experience has been gathered yet with respect to the recently proposed hirudin [1].

Blood smears colored by an optimized Pappenheim staining (May-Grünwald-Giemsa) [2] were scanned in the fully mechanized microscopes and the digital images then automatically analyzed. For this purpose, the slides are automatically or manually oiled, the valid range determined at a low magnification, the position of the white blood cells identified, red blood cells photographed and then the leukocytes photographed at high magnification individually and classified. The number of slides that can be examined in a single pass is limited, depending on the model, to 8, 32 or 96 slides. The older systems of CellaVision and the current HemaCAM contain still recognizable complete microscopes, while the traditional microscope form cannot be found in newer systems, since the beam path was integrated into individual parts and the unit mechanized in a much more compact form, doing away with eyepieces and visual observation.

Critical steps in the process of analysis of a specimen slide are the detection of the valid range, in which a differentiation is possible (individually positioned erythrocytes) and the detection of a single leukocyte. The necessary algorithm for the separation of adjacent cells is not error-free and requires in both instruments some reworking by the user. The classification of leukocytes includes the known subgroups of granulocytes with basophils, eosinophils and neutrophils, including their unsegmented precursors, metamyelocytes, myelocytes, promyelocytes and myeloblasts. Among lymphocytes, one device differentiates activated forms.

The DM software is based on the basis of so-called neural networks that operate non-hierarchically and whose decisions cannot be programmed. The HemaCAM software is based on a proprietary hierarchical classification model and is trained with data sets of known cell assignment. This has been prepared by the company and is optimized for a specific coloration that needs to be reproduced.

As hematology devices, CellDyn 3500 and CellDyn Sapphire were used (both from Abbott Diagnostics), and for immunophenotyping, the flow cytometer Cytomics FC500 from Beckman Coulter. For immunological differentiation, the following fluorescently labeled monoclonal antibodies were used: CD3, CD4, CD8, CD14, CD16, CD19, CD45, CD56, CD123 and HLA-DR. Leukocytes were defined as CD45-positive events, wherein by combining immunofluorescence with light scattering analysis leukocyte aggregates, platelet aggregates, macro platelets and erythrocytes were excluded from the analysis.

Neutrophilic granulocytes were assessed by a combination of CD16, CD45, and side scattered light. Eosinophils have a higher expression of CD45, a higher side scatter and are negative for CD16. Immature granulocytes have a lower or absent CD16 expression, with a contamination with monocytes by the below-described antibody combination having been safely ruled out. Basophils can be detected reliably and differentiated well from CD123+ dendritic cells due to the negativity for HLA-DR on CD123+ dendritic cells.

Monocytes were defined immunologically as cells with a high CD45 expression, low CD4 density and lack of CD3 marking, and with simultaneous HLA-DR expression, since CD14 alone does not cover all monocytes.

Lymphocytes were defined as the sum of T-cells (CD3+), B-cells (CD19+) and NK-cells (CD16&56+ and CD3-). The determination of CD4 and CD8 on T-cells was not taken into account apart from the fact that the lymphocyte gate was intentionally so broad that even granular lymphocytes were largely detected.

Results

Possible evaluations and comparisons of the data are:

  1. Comparison of the automated with the manual evaluation of digital images, i.e., the same cells are rated

  2. Comparison of the automated differential blood count with a separately obtained manual differential blood count (same slides, different cells)

  3. Comparison of the automated differential blood count with the mechanical differential blood count

  4. Comparison of the automated differential blood count with the immunocytometric differential blood count.

If you do this for all subclasses, a wealth of comparisons is generated, which is why we focused on primary examples for the purposes of this publication.

In the statistic evaluation, many comparative studies in other publications tend to draw upon the Spearman’s rank correlation, including the studies referenced in the literature with respect to the CellaVision devices. Individual high values force the line and determine the slope of the linear equation. Therefore in this publication, for each comparison the Bland-Altman plots were shown that reveal the true extent of the variance, particularly in the lower range of values.

Thrombocytes

Given the weaker magnification and resolution, thrombocytes can hardly be assessed in the overview with a 10×objective. With regard to the platelet morphology, the devices detect macro platelets that are found in the search for leukocytes and classified. Platelet aggregates are not yet recognized by the two devices examined for this study. The full investigation of thrombocytopenia is not yet possible in this way, and therefore requires further work at a regular microscope by a technical lab assistant. Company representatives for either equipment did promise that this would be corrected. So far, thrombocytes are photographed only in overview images together with the erythrocytes, but the low magnification and resolution are not sufficient for assessing the granulomere (e.g., dysplasia of platelets). But macro platelets are then often identified as leukocytes and discarded during classification. Here one finds a surprisingly good correlation for manual post-analysis (Figure 4).

Figure 4 to 15 Representation of the correlation diagrams of automated and manual recognition of cells with automated microscopy for the most important types of leukocytes based on the same data set of digital images.(A) and (B) x-y and Bland-Altman plot concerning macro platelet recognition (=giant Plt=macro Plt) on DM1200, automated vs. post-adjusted.
Figure 4 to 15

Representation of the correlation diagrams of automated and manual recognition of cells with automated microscopy for the most important types of leukocytes based on the same data set of digital images.

(A) and (B) x-y and Bland-Altman plot concerning macro platelet recognition (=giant Plt=macro Plt) on DM1200, automated vs. post-adjusted.

Erythrocytes

For the assessment of erythrocytes, visual fields of 10–20,000 erythrocytes are offered, and it is only the DM1200 that automatically determines microcytosis, macrocytosis and anisocytosis of erythrocytes. Shape changes or inclusions (Howell-Jolly bodies, etc.) will not be analyzed, that is, the detection of morphological changes such as fragmentocytes, tear shapes or target cells for clarifying a cause of anemia based on the changes to red blood cells is not yet possible. But fragmentocytes, for example, are an indicator of a series of diagnoses [3].

The assessment of the erythrocytes is carried out in the low magnification and resolution, which is used to search for leukocytes. In this overview, which includes a few thousand erythrocytes, frequent and gross changes in size and shape are also included. However, the size and anisocytosis are already known from the hematology device. The additional detection of poikilocytosis is reliably possible, but only by purely manual, and not automated, assessment. Rarer and diagnostically significant shape changes such as fragmentocytes can escape observation or their counting requires an examination of a larger number of erythrocytes. Polychromasia is difficult to detect. The same is true of any inclusions such as basophilic stippling, Pappenheim’s bodies, Cabot rings or parasites. If these happen to be in the neighborhood of a leukocyte, they are photographed at a high magnification, and with careful observation they would be detected by the user, but not recognized in the overview or through automated means.

The really obvious digital assessment of a reticulocyte stain with vital dyes (brilliant cresyl or new methylene blue) is not provided for with any of the systems.

Nucleated red blood cells (erythroblasts) are detected in the leukocyte analysis and recognized as such reliably. Quantification with low occurrence is subject to the same statistical restrictions as manual microscopy. A subdivision according to maturity or the recognition of a nuclear dysplasia is not provided.

Leukocytes

For all devices, the automated differentiation of normal leukocytes is in the foreground, which is why only this was compared to other methods. But even these cells cannot be classified completely as a rule, and there remains a residue of 2%–10% of leukocytes per sample that must be assigned or classified manually by the observer. The more abnormal cells, the higher the number of cells that need to be classified subsequently. The primary design goal for both units was the correct allocation of normal white cells of the blood count.

The demands on the morphological integrity of the leukocytes are high: Aging leukocytes from externally submitted, unstained blood samples, which may be one or more days old, with aging artifacts such as granulocytes are unsuitable. However, the systems are also faced with such samples. We ourselves also included samples that traveled 24 h (temperature-stabilized express delivery at 10–14°C).

The respective statistical results of the comparisons of the two systems with conventional methods are listed in Tables 1 and 2. The figures show the correlation between the device and observer on the same imagery (HemaCAM Figures 5–10, DM1200 Figures 11–22).

Figure 5 (A) and (B) x-y and Bland-Altman plot for neutrophilic granulocytes (manual vs. HemaCAM).
Figure 5

(A) and (B) x-y and Bland-Altman plot for neutrophilic granulocytes (manual vs. HemaCAM).

Figure 6 A and B: x-y and Bland-Altman plot for lymphocytes (manual vs. HemaCAM).
Figure 6

A and B: x-y and Bland-Altman plot for lymphocytes (manual vs. HemaCAM).

Figure 7 A and B: x-y and Bland-Altman plot for monocytes (manual vs. HemaCAM).
Figure 7

A and B: x-y and Bland-Altman plot for monocytes (manual vs. HemaCAM).

Figure 8 A and B: x-y and Bland-Altman plot for eosinophilic granulocytes (manual vs. HemaCAM).
Figure 8

A and B: x-y and Bland-Altman plot for eosinophilic granulocytes (manual vs. HemaCAM).

Figure 9 A and B: x-y and Bland-Altman plot for basophilic granulocytes (manual vs. HemaCAM).
Figure 9

A and B: x-y and Bland-Altman plot for basophilic granulocytes (manual vs. HemaCAM).

Figure 10 A and B: x-y and Bland-Altman plot for smudge cells (manual vs. HemaCAM).
Figure 10

A and B: x-y and Bland-Altman plot for smudge cells (manual vs. HemaCAM).

Figure 11 x-y and Bland-Altman plot for segmented neutrophils (manual vs. DM1200, x=DM1200, y=user).
Figure 11

x-y and Bland-Altman plot for segmented neutrophils (manual vs. DM1200, x=DM1200, y=user).

Figure 12 A and B: x-y and Bland-Altman plot showing the smudge cells as the cause of the difference in segmented neutrophils (user vs. DM1200).
Figure 12

A and B: x-y and Bland-Altman plot showing the smudge cells as the cause of the difference in segmented neutrophils (user vs. DM1200).

Figure 13 A and B: x-y and Bland-Altman plot for % nonsegmented neutrophils (Bands): x=DM1200, y=user.
Figure 13

A and B: x-y and Bland-Altman plot for % nonsegmented neutrophils (Bands): x=DM1200, y=user.

Figure 14 A and B x-y and Bland-Altman plot% lymphocytes: x=DM1200, y=user.
Figure 14

A and B x-y and Bland-Altman plot% lymphocytes: x=DM1200, y=user.

Figure 15 A and B: x-y and Bland-Altman plot% for reactive lymphocytes: x=DM1200, y=user.
Figure 15

A and B: x-y and Bland-Altman plot% for reactive lymphocytes: x=DM1200, y=user.

Figure 16 A and B: x-y and Bland-Altman plot for% blasts: x=DM1200, y=user.
Figure 16

A and B: x-y and Bland-Altman plot for% blasts: x=DM1200, y=user.

Figure 17 A and B: x-y and Bland-Altman plot for % monocytes: x=DM1200, y=user.
Figure 17

A and B: x-y and Bland-Altman plot for % monocytes: x=DM1200, y=user.

Figure 18 A and B: x-y and Bland-Altman plot for % monocytes, user DM1200 vs. manual differentiation. Please compare with Figure 17, where the same images (cells) were classified.
Figure 18

A and B: x-y and Bland-Altman plot for % monocytes, user DM1200 vs. manual differentiation. Please compare with Figure 17, where the same images (cells) were classified.

Figure 19 A and B: x-y and Bland-Altman plot% monocytes, user DM1200 vs. immunodiff. Please compare with Figure 17. The worse correlation in Figure 19 shows that the uneven spatial distribution of monocytes in a wedge blood film is the major source of error.
Figure 19

A and B: x-y and Bland-Altman plot% monocytes, user DM1200 vs. immunodiff. Please compare with Figure 17. The worse correlation in Figure 19 shows that the uneven spatial distribution of monocytes in a wedge blood film is the major source of error.

Figure 20 A and B x-y and Bland-Altman plot% monocytes, CellDyn3500 vs. immunodiff.
Figure 20

A and B x-y and Bland-Altman plot% monocytes, CellDyn3500 vs. immunodiff.

Figure 21 A and B x-y and Bland-Altman plot% eosinophils: x=DM1200, y=user.
Figure 21

A and B x-y and Bland-Altman plot% eosinophils: x=DM1200, y=user.

Figure 22 A and B x-y and Bland-Altman plot% basophils: x=DM1200, y=user.
Figure 22

A and B x-y and Bland-Altman plot% basophils: x=DM1200, y=user.

Table 1

Performance data of the HemaCAM system.

NeutrophilsLymphoc.Monoc.EosinophilsBasophilsPlateletsSmudge c.
Avg. share62.00%30.00%4.00%2.00%0.50%0.50%0.50%
no,valide326
Cells/Slide110
nRF*FN22233.2107581434.4717.2179.3179.3179.3
nRF*FP(TQ)22233.210125.21344.8675134.5179.3179.3
nRF*FP(SP)18169.133469.334425.646857.135680.735680.728544.6
nRN*FP13626.82510234425.635142.835680.835680.735680.7
Sensitivity
Specification85.0%80.0%75.0%80.0%60.0%80.0%75.0%
trQ2224107614472181818
nRP*FN15314562965143062129103
nTotal24700187641627521500124002580020600
nO,Total225171148196113235188
TQβ–q185.8%81.5%79.8%85.4%77.1%89.3%86.4%
Precision (Positive predictive value)
Specification85.0%85.0%80.0%85.0%80.0%80.0%75.0%
TGen(TQ)2224101313568141818
n–RP*FP(TQ)153147085768547106129103
nTotal(TQ)24700250932048029060282672580020600
nO,Total(TQ)225229187265257235188
Genβ–q1(TQ)85.8%86.2%84.1%89.2%90.1%89.3%86.4%
TGen(SP)1817334734434686356935692855
nRP*FP(SP)12555229081765431942182921829211730
nTotal(SP)24780245451839024446183841838414737
nO,Total(SP)226224168223168168134
Genβ–q1(TQ)85.9%85.7%80.9%85.6%80.9%80.9%76.2%
Specificity (=1–Failure rate)
Specification80.00%80.00%80.00%80.00%80.00%80.00%80.00%
tSP1363251134433515356935693569
nFP*RN7094129311765418019182921829218292
nTotal18669184731839018387183841838418384
nO,Total170168168168168168168
Spβ-0,181.4%81.0%80.9%80.9%80.9%80.9%80.9%
Table 2

Performance data of the DM1200 device.

Pre-classification data relativeShare of correctly pre-classified cellsShare of correctly pre-classified cells relative to verified cells
Cell classPre-classifying agreementIn agreement with final result
Segmented neutrophil (SN)99.3%97.2%
Eosinophil100.0%87.5%
Basophil86.4%100.0%
Lymphocyte98.6%100.0%
Monocyte100.0%98.4%
Unsegmented neutrophil (band, BN)52.2%87.8%
Variant lymph57.1%66.7%
Smudge cell99.8%95.1%
Erythroblast (NRBC)33.3%100.0%
Artifact99.3%92.1%
Macrothrombocyte87.3%97.0%
Pre-classifying agreementIn agreement with final result
Segm. Neutrophil (SN), Ly, Mo99.1%98.2%
SN, Ly, Mo, Bas, Band, Eos97.5%97.8%
Note. Variant lymphocytes were classified as “normal” in the subsequent calculations
%%
Abnormal designated as normal0.7%0.3%
Normal misclassified in another normal cell class1.2%
Normal designated as abnormal0.2%0.1%

Pre-classified in agreement: These are all WBC and non-WBC that are in agreement with the pre-classification of the DM following validation. Accuracy (all objects) 96%: This is the accuracy of the DM following validation by Henk J (this is the most important parameter of the spreadsheet). It means that on average, 4 cells of a differential blood count of 100 cells were shifted to another class. Accuracy (when SN=BN) 97%: This is the accuracy of the DM following validation by Henk Jansen when “band cells” (segmented, BN) are classified as segmented cells (SN). Number of re-classified objects: The cells sorted into cell classes that are not classified automatically by the DM [e.g., large granular lymphocytes (LGL cells)]. Pre-classification data relative: For all cell classes, accuracy of classification is computed as pre-classification accuracy, e.g., for segmented, it is 99.3%. SN, Ly Mo Segmented neutrophils+lymphocytes+monocytes have 99.1% accuracy. SN, Ly, Mo, Bas, Band, Eos The six normal cell classes have a 97.5% accuracy (note that this includes unsegmented).

Neutrophilic granulocytes can be reliably detected (Figure 5 or 23), however, the distinction between segmented and unsegmented according to the so-called rule of thirds, where the definition of unsegmented should be based on the minimum core diameter being one-third or more of the maximum, is rather poor with both systems (HemaCAM not in the specifications, DM1200 Figure 12). It requires post-processing for each sample, which can negate the time saved with automation (work relief). DM1200 counts, by approximately a factor of two, more unsegmented than an expert, with the results being correlated poorly. The separation of segmented and unsegmented neutrophils or lymphocytes and LGL cells is not realized satisfactorily by either device, which would also apply to the comparison of two investigators. The poor separation of unsegmented and segmented was reported in a previous study [4]. The DGHO Laboratory working group, given the poor reproducibility, has also questioned the meaningfulness of such distinction even in the case of a human observer (publication with recommendations for differential counts in preparation).

Figure 23 A and B: x-y and Bland-Altman plot for % basophils, user DM1200 vs. immunodiff.
Figure 23

A and B: x-y and Bland-Altman plot for % basophils, user DM1200 vs. immunodiff.

Lymphocytes are the most demanding leukocyte subpopulation, in addition to the blasts, which is why the working group has published a guideline to harmonize the nomenclature and proposed an iterative process for the classification [5]. In the selecting of images for standard and reactive lymphocytes, it became clear that the transitions are fluid and even the co-authors use different boundaries and tolerance levels. Software cannot solve this problem and the training data set for the cell classification is characterized by the person who has defined the cells therein. The correlation for standard lymphocytes is reasonably good (Figure 6 or 13). The one for the reactive forms is poor (Figure 14).

The LGL cells (large granular lymphocytes) are not currently classified (or shown separately) by either company. There are two main types: those with few (about 3–7) and relatively coarse and those with more (7–15) and somewhat finer granules. The former, according to own experience, are NK cells (CD16&56+, CD3-) while the second group rather corresponds to the NK-like T-cells (CD3+56+) (T. Nebe, data not published). The latter cells are overlooked by an inexperienced observer at first glance. In patients with LGL diseases in this category, one can observe over time fluctuating values mainly caused by different examiners, although immunologically determined quantities are usually very consistent. In the subsequent differentiation of these smears by a single person, the results became fairly constant (T. Nebe, data not published). The interest in LGL mainly arises from the clarification of unclear neutropenia, where LGL-induced neutropenia is more common than one may think. Rheumatoid arthritis is often associated with LGL proliferation, which in no way constitutes a specific feature, but may be associated with neutropenia that might also be drug dependend. LGL cells should therefore be included in future specifications.

The eosinophils and basophils can be seen by the human examiner clearly and easily, which is why the poor recognition of these cells by the devices creates a lack of confidence. Above 3% (DM1200) or 5% (HemaCAM) eosinophils are better classified (Figure 8 or 20). The basophils are fairly reliably detected only by DM (Figures 9 and 21). Comparing the tested immunological basophil count, one finds that the DM1200, approximately by a factor of two, measures higher percentages of basophils in terms of WBC, but the correlation with the immunologically determined proportion is very poor (Figure 22). A possible explanation would be, given the inhomogeneity of the spatial distribution with accumulation at the center of the slide, the omission of the marginal ranges in the mechanical scanning method (DM1200). When a comparison is made with another slide or cytometric method, the correlation diagram shows the inevitable scattering at low percentages and low amounts of cells counted (sum counted n=100 or 200). Generally, the VK is halved only if the sum counted is quadrupled (square root in the VK formula). Up to 400 cells can be differentiated in the peripheral smear even without reaching areas of too dense or too thin red cells at the end of the smear.

The monocytes, because of their diversity, have always been the problem child in the differentiation of the blood smear and hematology analyzers. Under 10%, i.e., in the normal range, the detection yields useful results only with the DM system (Figures 7 and 16). For demonstration purposes, we also showed here comparisons with other methods. Although the correlation machine verus man for the same images (!) is good, it is poor for the immunological differential blood count (Figure 19). The slide methods do not correlate well with each other (DM1200 vs. manual differentiation, same slide but assessment of different areas) (Figure 9), as already shown by Ceeli [6].

Smudge cells

On the recommendation of the DGHO Laboratory working group, the smudge cells are included in the 100% leukocytes of the differential count, so that the clinically relevant absolute values of the leukocyte subsets are estimated correctly [7]. An extreme example is the smudge cells in chronic lymphocytic leukemia, where often one-third of leukocytes are smudge cells and, in case of their omission and leukocytosis, the percentage of neutrophils is overrated, creating excessive values in the calculation of the absolute neutrophil count. Therefore, this approach has prevailed in the German speaking countries and is also being studied as part of interlaboratory testing. The HemaCAM system takes the smudge cells into account in accordance with this recommendation. But smudge cells also occur frequently with other leukemias, viral infections and aged blood samples. Even a normal blood smear of a healthy person contains a few smudge cells. They are often ignored and no longer perceived in the long run, and then not integrated into the differential count. The type of pre-cleaning of the slides supplied also plays a role and must be taken into account. The smudge cells tend to explain best the discrepancies between segmented from machine vs. man, as Figures 11 and 12 show.

In the HemaCAM automated microscopy and the manual post-assessment of the automated differential blood counts, smudge cells were taken into account, but not by the CellaVision system, where they were not considered as leukocytes. A correction of this problem has been promised.

Blasts

The term “blast” is not standardized internationally. This publication defines it as an immature blast of an AML, of a transformed myeloproliferative disease or as a lymphoblast of an ALL or a highly malignant lymphoma. Circulating progenitor cells within the meaning of myeloblasts would morphologically also be included as blasts, since they are not clearly distinguishable from malignant (clonal) blasts in the Pappenheim staining.

The comparison was made only for the DM system, as an automated blast allocation was specified only for this system. Below 4%, the value is completely unreliable. The blasts are overestimated (Figure 15) and therefore the values are useless. Cornet and colleagues confirm that a correct assignment to the blast types such as AML is not achieved [8].

Digital image quality

The resolution of the digital cameras does not represent the sole limiting factor, as even the resolution of a 5 MP camera is above that of the 100×lens; with the 40×lens, however, this is limiting. But true colors and contrasts play an equally important role. The differences in quality of two camera chips with the same number of pixels are well known from the practice of normal photography. The software (database) for cell classification is usually initially trained by the manufacturer, which is why his color quality requirements may not match the user’s ones (saturation, color balance). In future systems, a 63× microscope objective should be used, instead of the 100×lens used by both companies currently, because the 63× has a higher numerical aperture (resolution) and a greater depth of field compared to the 100×lens. But the digital camera is chosen at the beginning of the device development, long before the market launch, and then kept, so that when it is released, it is no longer state-of-the-art, with the younger system having a better image quality, which also depends on the new illumination by means of a good white LED. With development times of five or more years, it is a substantial drawback that the camera cannot be replaced, particularly in view of the rapid development of the photo chips.

In hematology, special attention has to be given to problem smears or patients, with high demands on resolution (azurophilic granules of monocytes, Auer rods, etc.) or on the search for rare cells e.g., blasts in connection with MDS or relapse, or minimal myoloid differentiation of immature blasts in acute leukemias. In these cases, the benefit of a sole assessment of digital images and, thus, a remote assessment of difficult smears remains questionable, and it should be noted that the screen resolution and the duration of transmission of the images via the internet represent limiting factors. This becomes immediately apparent in experiments with telemicroscopy on complex samples, such as bone marrow smears. The quality of digitized images in combination with a good lens and a numerical aperture of 1.4 is still below the resolution of the human eye looking directly through a microscope, which is caused by the limited pixel resolution (per pixel several pixels for separate colors), color depth, true colors (many cameras produce bluish images) and contrast of digital cameras. The current standard color depth of 8 bits of digital cameras limits the assessment of the more or less violet nuclear chromatin, where a high spread of brightness and colors would be necessary in a narrow range.

Software ergonomics

The usability of the software plays an essential role, not least when considering the time savings. The clarity of presentation, especially the contents of the cell classes, the correction of cell classification, the parallel evaluation of the measurement during the scanning process, quality control and on-line communication support with the laboratory information system (HL7 interface, ASTM protocol, image transfer), and the presentation of the results from the hematology analyzer to validate the results must be observed. Also critical are the number of mouse clicks, keyboard shortcuts for commands, a batch mode for processing serial measurements, telemicroscopy support (consulting), external support by remote access, to name but a few aspects. Here the standards that already exist for routine clinical laboratories are not yet fully met.

Discussion

Two of the most highly developed systems for digital microscopy of blood smears have been studied in order to take a critical look at fundamental questions on the use of this technology in clinical routine practice to prepare a differential blood count. Therefore, smudge cells and samples no more than a day old were included (temperature-stabilized transport) were included in the trial, which resulted in poorer correlations compared to previous studies.

Recent decades have seen a change in the role of the hematological differential blood count derived from blood smears plays: While in the past the focus was on differentiatial count of the five leukocyte subgroups, because the mechanical device supplied only the leukocyte count, this result in todays laboratory is largely done by hematological analyzers based on the principle of flow cytometry. Apart from monocytes and basophils (depending on the differentiation principle), they produce a relatively reliable normal five-part differential and blood count. The real challenge for technical lab assistants nowadays is to recognize abnormalities of cells, such as pathological erythrocyte morphology, dysplasia, blasts, lymphoma or LGL cells, i.e., balancing the hematological information deficit of modern hematology equipment. The pathognomonic changes of erythrocytes, especially poikilocytosis of erythrocytes with special shapes such fragmentocytes or inclusions, are indispensable in the clarification of anemia (Table 3). In terms of a more specialized investigation of the leukocytes, this is followed by immunophenotyping with the so-called immunological differential blood count. Here, through simultaneous staining with several monoclonal antibodies with different fluorescent labeling and multi-parametric evaluation, a differentiation is carried out. The pending standardization and automation are still holding up its routine use – in addition to the higher costs. Table 3 shows, therefore, a comparison of the two systems studied, with the hematology instruments as well as manual microscopy and immunocytometry.

Table 3

Detection of abnormalities in blood count as evaluation criteria for automated blood microscopy.

Table 3 Detection of abnormalities in blood count as evaluation criteria for automated blood microscopy.

The assignment of normal leukocyte subclasses of healthy blood counts is relatively well feasible with both systems, but in any case requires a more or less substantial readjustment. Whether this involves 2, 5 or 10 cells is secondary, because primarily the work flow is disturbed. In the reactive left shift of neutrophilic granulocytes, differentiating the unsegmented from the segmented ones does not work very well on either system and probably cannot be sustained as a specification. It does raise the question to what extent this poorly reproducible parameter is still clinically relevant in these days and age of the C-reactive protein (CRP) and procalcitonin (PCT). Another problem is the differentiation of reactive lymphocytes from monocytes, and the detection of small monocytes. Even large granular lymphocytes (LGL cells) cannot be distinguished by the devices from other lymphocytes, because the systems were not trained for this. The recognition of blasts succeeds only partially, as shown by Billard et al. too [9].

The post-assessment of leukocytes not or incorrectly allocated, thus, requires a correction on the device screen. But the insufficient or absent evaluability of blasts, erythrocytes, thrombocytes and rare cells often requires both, i.e., a follow-up examination at the regular microscope. Depending on the specimen material (ambulatory persons, hematological patients or intensive care patients), the automated classification by the device is more or less successful and requires more or less extensive post-processing.

It is thus mainly laboratories that benefit from computer-aided microscopy, which receive many normal samples of healthy individuals and whose hematology analyzers wrongly attach warnings to many normal samples (false-positive). Based on the question (initial diagnosis or follow-up monitoring of hematological system diseases) and the type of atypical characteristics, one must decide, on the basis of the digital images, whether subsequent traditional microscopy will be necessary. The time required for subsequent adjustments on the PC and follow-up by means of conventional microscopy, thus, represents the essential factor in estimating personnel input (aside from the low maintenance requirements of the systems).

(In-) Accuracy

What is the true result for each cell class? Comparisons of microscopic with mechanical and immunological differential blood counts are not new. The primary comparisons made here in the present study are based on the differentiation of the same slide, i.e., man versus machine, with respect to the same totality of digital images. The images were not fundamentally reclassified, but re-assessed after the cells had already been pre-classified by the system. This means that the results of such comparisons are biased by prejudice. Comparing an automated differential blood count to microscopic direct evaluation of the same smear, this becomes clear due to a worse correlation, because not the identical population of cells was observed. Any good hematological laboratory is aware of the differences that arise when two different people count out the same slide. The wider variances of manual methods in comparisons with hematology analyzers are first of all attributable to the statistics of fewer cells (n=100 or 200 under the microscope vs. n=10,000 in the hematology analyzer). The dependence of the probability on the number and frequency of the cells counted was already described by Rümke in 1960 [10], and even automated microscopes cannot bypass that basic rule.

Often, however, the inhomogeneity of the smear is overlooked as one of the major causes. There are studies carried out decades ago that show different probabilities of cells occurring in different areas of the slide [11]. Thus, activated lymphocytes or monocytes are more likely to be located at the end of the peripheral smear or the edge of the smear; blasts tend to be found at the edge and basophils in the center of the smear. The NCCLS guidelines (now called CLSI) for the homogeneous cell distribution called as the reference method for the centrifugation technique, but this has not been adopted so far in practice in any country [12]. The comparisons DM1200 vs. manual differentiation and DM1200 vs. hematology unit impressively demonstrate the large variance as a result of statistics and smear technique. No matter how qualified the morphologist might be, he or she cannot beat this inhomogeneity and the statistics in counting out. The deviation from the true value in the blood in the case of blood smears, therefore, cannot be blamed solely on digital image analysis. Then, there are the smudge cells, whose origin is not always detectable and which are not to be assigned to any cell series.

Therefore, the classic blood smear cannot be the gold standard for the differential count and thus the determination of the concentration of leukocyte subpopulations in the blood, but it can be that for the morphological identification of pathological conditions.

Conversely, the hematology machines, especially in reactive and malignant lesions, cannot distinguish sharply the leukocyte subsets, e.g., between reactive lymphocytes and monocytes, and rare cells such as basophils or individual blasts often go unnoticed in the mass (neutrophils, lymphocytes). The immunological method, i.e., the flow cytometric immunophenotyping with fluorescent monoclonal antibodies, uses in the default implementation lysis reagents for the removal of erythrocytes. Lysis reagents of varying compositions lead to systematic differences in the differential blood count, as each laboratory can easily verify in validating its methods. Only the examination of diluted unlysed whole blood using immunophenotyping is the gold standard in this context, and is also used by the Physikalisch-Technische Bundesanstalt (PTB – Germany’s national test authority) in Berlin as the reference method for interlaboratory testing for differential blood counts to verify the target values. Nevertheless, not all immunocytometric values can be applied on a one-to-one basis, since, for example, the left shift is defined and broken down in morphologically and immunologically different ways. For rare cells such as blasts, basophils or progenitor cells, immunophenotyping is a true enhancement.

Economic aspects

The cost of purchasing automated differential microscopes is equivalent to one to two years’ salary of a technical lab assistant (including non-wage labor costs) and would amortize in one to three years from the perspective of the finance department if staff retention were greatly reduced. Based on such estimate, the acquisition of such systems is frequently welcomed and supported by hospital or laboratory management.

The authors with hematological experience, who have evaluated these systems so far, agree in their assessment that in the case of hematological patients a longer processing time will be required than without the system and that also in the case of reactive blood samples, the savings compared to the conventional working method will be insubstantial. Surcouf and colleagues, however, have reported a time saving of 30% [13]. Ceelie and colleagues have identified a time saving with the DM96, a system faster than the DM1200, in connection with non-pathological samples of 1 min./slide (3.3 min./slide with manual differentiation of 100 cells) in favor of more time for manual analysis [6]. Kratz and colleagues see time savings only in connection with less experienced workers [3]. The extent to which working hours can be reduced with normal blood counts is also dependent on the individual laboratory, that is, in terms of the patient population and upstream hematology system that sorts the samples more or less well that, then, are no longer supposed to be handed on to computer-aided microscopy. In all this, the set rules defined by the laboratory play an important role, as they set out the criteria for subsequent differentiation via conventional microscopy. As experience has shown, the limits for patients from hematological departments in connection with blasts (relapse) are stricter than for ICUs with septic patients. This manual post-examination at the regular microscope, including locating the slide again, has substantial adverse effects on time savings. The manufacturers of the two microscopy systems discussed here, when asked explicitly, refused to say that the critical and consistent application of such rules can generate substantial savings with respect to staff positions by means of assisted microscopy. However, better standardization, compared to the manual method, can surely be expected among different investigators at the same laboratory.

Telemedicine

The regular electronic image documentation of all scanned cells also allows for remote peer review internally or at remote locations and possibly for the images to be included with the findings, which is an important argument in favor of such systems. Based on experience with a share of hematological patients of more than 50%, the sole assessment of the digital images, is sufficient only in connection with inflammatory changes and relapse-free follow-up testing, is sufficient. For each initial anemia work up involving the question of erythrocyte morphology, for each new case of thrombocytopenia to rule out aggregates, and when looking for blasts at the edge of the smear, conventional microscopy is still necessary. Another aspect of post-assessment is the better optical quality of the images obtained through a conventional microscope (subject to sophisticated lenses and lighting).

Digital images can be discussed in a video conference either on the basis of still images (usually compressed in JPEG, about 1 MB in size) or, preferably, in the form of live recordings, where the bandwidth of the Internet connection poses the only potential limit. The possible resolution (>5 MP) of digital still cameras with live images cannot be broadcast yet live with attractive frame rates. Our own experience with video conferencing involving a discussion of live images and using conventional microscopy via digital cameras shows that the experts obtain a lot of information from the context of the sample or patient to compensate for the poor resolution and color fidelity. As a rule, three to four difficult cases are discussed in 30–45 min. The continuous increase in the bandwidth of Internet connections and the storage capacity of digital images support these efforts. However, these objectives cannot be achieved in the short term with the devices discussed herein, which is also why some laboratories have so far decided against purchasing such equipment. Sending digital images by e-mail, followed by a discussion by written correspondence, is more time-consuming, and the contextual information from the discussion is missing.

Other applications

The automated analysis of cytological cell images, which are more complex than a blood smear in terms of the cell types and the juxtaposition of cells, as is true of cytospin preparations, cervical swabs or bone marrow, is not yet possible with the aforementioned equipment. Along these lines, therefore, the manufacturers should primarily consider image digitization by way of an image scanner. But the optical quality of the weaker magnifications used thus far is not sufficient for this purpose. Still, this application is among the stated objectives for the further development of the devices, according to both manufacturers and new systems are coming up. What is more, the required storage space and the time for scanning the slides with these applications are considerably greater than when photographing 200–400 individual cells in the blood smear. The examination of urine sediments by means of phase contrast microscopy would be a further possible and rewarding task for such systems.

Conclusion

The conclusion remains that the current systems of computer-aided microscopy for the differentiation of blood smears now allow for a smooth operation, while automatic image recognition for normal blood samples yields a good, but not 100% correct, allocation of the leukocytes by the software. The true values in the patient’s blood will be hard to determine with a smear method per se. Digital images are available for remote viewing, which is supported more or less well by the current software. On a positive note, the accuracy of the allocation of cells is, at best, above that of a technical lab assistant with little experience in the field. But it is precisely such an assistant who cannot identify and correct all of the cells misallocated by the system, which is why the systems must be operated by people experienced in blood cell differentiation. The decision on whether the allocation made by the system is correct, thus, requires a person especially experienced in hematology who must be able to manage with a poorer digital image quality relative to direct visual observation. This is in contrast to the frequent desire of the buyer or user to let hematologically less experienced staff operate the device – and with fewer working hours. In addition, the automated analysis of erythrocytes and thrombocytes is currently limited, which, from a hematological point of view, represents an obvious design flaw on the part of the device developers. Anemia is the most common pathological condition in the blood count, which means the analysis of erythrocytes (and reticulocytes) ought to be supported better. The outcome of a primary assessment of leukocytes is, thus, a flawed fundamental assumption due to the lack of hematological and/or clinical experience – and this the background to current attempts to correct and improve the devices.

A personal discussion with colleagues at larger private labs, which analyze up to 10,000 blood samples a day and which use such systems, shows that the devices became necessary mainly on account of the work volumes to be processed.

Especially in such laboratories, a large number of normal and minimally reactive blood counts accumulate, which are ordered by primary physicians for the purposes of screening or clarification of unspecific clinical symptoms. Genuine pathologies are rarely found here. Even minor non-specificities of blood count analyzers then lead to enormous amounts of smears, which must be differentiated subsequently but do not exhibit any or only irrelevant abnormalities. This can, indeed, save time, as described by Ceelie [6].

If this time saving is used to identify the actual conspicuous individual sample and to subject it to a thorough critical analysis and subsequent correction at the microscope, the use of an automated differentiation system may actually help improve the quality.

Otherwise, a critical analysis and subsequent correction of the individual sample are impossible in practice in view of the remaining high number of samples (Figure 1), despite a diagnostic screening. The reimbursement paid for a microscopic differential blood count under the billing system of the statutory health insurance companies in Germany, which is EUR 0.40 (!!), does not even come close to covering the cost, not even with these new systems. Conversely, the above situation shows the complex detection and decision processes behind the differentiation of blood smears and the experience that is necessary, but completely ignored by the remuneration systems and substantially underestimated by a laboratory management with very little experience in such things that often decides to purchase such equipment.

No doubt both systems discussed have an educational character, i.e., a new and inexperienced worker is given pre-classified data sets. Then, one’s own result can be compared with the target values and/or expertise using special software, thus identifying the deficits. Therefore, use for server-based digital interlaboratory testing as part of quality assurance is being prepared as well. For such purpose, the data set of an entire scanned slide sould be available in high resolution to prevent any limiting pre-selection of cells.

Correspondence: Carl Thomas Nebe, Hämatologie-Labor Mannheim, Collinistr.11, 68161 Mannheim, Germany, Tel: +49-621-43 73 29 91, Fax: +49-621-43 73 67 33
  1. 1)

    Original German online version at http://www.degruyter.com/view/j/labm.2012.36.issue-5/labmed-2011-0036/labmed-2011-0036.xml?format=INT. The German article was translated by Compuscript Ltd. and authorized by the authors.

We are grateful for the qualified assistance of Ms. Michaela Neuske and Ms. Anja Praß-Stähler in the identification of the differential blood counts.

Conflict of interest statement

Authors’ conflict of interest disclosure: The authors TN, MO and GS stated that there are no conflicts of interest regarding the publication of this article. HJ is an employee of Sysmex and TZ is the developer of HemaCAM.

Research funding: None declared.

Employment or leadership: HJ is an employee of Sysmex and TZ is the developer of HemaCAM

Honorarium: The authors TN, MO and GS stated that they received no honorarium regarding the publication of this article or by other means.

References

1. Menssen HD, Melber K, Brandt N, Thiel E. The use of hirudin as universal anticoagulant in haematology, clinical chemistry and blood grouping. Clin Chem Lab Med 2001;39:1267–77.10.1515/CCLM.2001.204Search in Google Scholar PubMed

2. Binder T, Diem H, Fuchs R, Gutensohn K. Pappenheim stain: description of a hematological standard stain. History, chemistry, procedure, artifacts and problem solutions. J Lab Med 2012;36: 293–309.Search in Google Scholar

3. Diem H, Binder T, Heimpel H. Fragmentozyten im peripheren Blut. J Lab Med 2005;29:331–2.Search in Google Scholar

4. Kratz A, Bengtsson HI, Casey JE, Keefe JM, Beatrice GH, Grzybek DY, et al. Performance evaluation of the CellaVision DM96 system: WBC differentials by automated digital image analysis supported by an artificial neural network. Am J Clin Pathol 2005;124:770–81.10.1309/XMB9K0J41LHLATAYSearch in Google Scholar

5. Baurmann H, Bettelheim P, Diem H, Gassmann W, Nebe T. Lymphozytenmorphologie im Blutausstrich – Vorstellung einer überarbeiteten Nomenklatur und Systematik. J Lab Med 2011;35:261–70.Search in Google Scholar

6. Ceelie H, Dinkelaar RB, van Gelder W. Examination of peripheral blood films using automated microscopy; evaluation of Diffmaster Octavia and Cellavision DM96. J Clin Pathol 2007;60:72–9.10.1136/jcp.2005.035402Search in Google Scholar PubMed PubMed Central

7. Diem H, Binder T, Bettelheim P. Kernschatten im Blutausstrich. J Lab Med 2005;29:333–4.Search in Google Scholar

8. 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;30:536–42.10.1111/j.1751-553X.2007.00996.xSearch in Google Scholar PubMed PubMed Central

9. Billard M, Lainey E, Armoogum P, Alberti C, Fenneteau O, Da Costa L. Evaluation of the CellaVision DM automated microscope in pediatrics. Int J Lab Hematol 2010;32:530–8.10.1111/j.1751-553X.2009.01219.xSearch in Google Scholar PubMed

10. Rümke CL. Die Zelldifferenzierung von Blutausstrichen: Variabilität der Ergebnisse, Triangel, 1959;4;154–8.Search in Google Scholar

11. Wolf I, Boroviczény KG. Probleme der Ausstrichdifferenzierung. In: Boll I, Heller S, editors. Praktische Blutzelldiagnostik. Berlin: Springer, 1991:387–402.Search in Google Scholar

12. National Committee for Clinical Laboratory Standards (NCCLS). Reference leukocyte differential count (proportional) and evaluation of instrumental methods: approved Standard. [.NCCLS document H20-A (ISBN 1-56238-131-8)Villanova, PA:NCCLS, 1992 und Clinical and Laboratory Standards Institute (CLSI). Reference Leukocyte (WBC) Differential Count (Proportional) and Evaluation of Instrumental Methods; Approved Standard- Second Edition, H20-A2, 2nd edition.2010.Search in Google Scholar

13. Surcouf C, Delaune D, Samson T, Foissaud V. Automated cell recognition in hematology: CellaVision DM96 TM system. Ann Biol Clin (Paris) 2009;67:419–24Search in Google Scholar

Published Online: 2013-09-04

©2013 by Walter de Gruyter Berlin Boston

This article is distributed under the terms of the Creative Commons Attribution Non-Commercial License, which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

Articles in the same Issue

  1. Masthead
  2. Masthead
  3. Editorial/Leitartikel
  4. Editorial/Leitartikel
  5. Molekulargenetische und zytogenetische Diagnostik/Molecular-Genetic and Cytogenetic Diagnostics
  6. Zellfreie fetale DNA im mütterlichen Blut: neue Möglichkeiten in der pränatalen Diagnostik/Cell free fetal DNA in maternal blood: new possibilities in prenatal diagnosis
  7. Cell-free fetal DNA in maternal blood: new possibilities in prenatal diagnostics1)
  8. Non-invasive prenatal testing for whole chromosome abnormalities/Nicht invasives pränatales Testen auf Chromosomenstörungen
  9. Non-invasive prenatal testing for fetal aneuploidy by massively parallel DNA sequencing of maternal plasma: the future has arrived today/Nicht-invasiver Pränatal-Test auf fetale Aneuploidie mittels massiv-paralleler DNA-Sequenzanalyse im mütterlichen Plasma: in der Zukunft angekommen
  10. Molekulare Karyotypisierung in der Routinediagnostik – Rückblick und Ausblick/Molecular karyotyping in routine diagnostics – a view back and forth
  11. Molecular karyotyping in routine diagnostics – a view back and forth1)
  12. Klinische Chemie und Stoffwechsel/Clinical Chemistry and Metabolism
  13. Vitamin D und kardiovaskuläre Erkrankungen: Standortbestimmung 2012/Vitamin D and cardiovascular diseases: where do we stand in 2012?
  14. Editorial/Leitartikel
  15. 100 Jahre Pappenheimfärbung
  16. Hämatologie/Hematolgoy
  17. Pappenheim-Färbung: Beschreibung einer hämatologischen Standardfärbung – Geschichte, Chemie, Durchführung, Artefakte und Problemlösungen/Pappenheim Stain: Description of a hematological standard stain – history, chemistry, procedure, artifacts and problem solutions
  18. Evaluation von mikroskopischen Systemen zur Erstellung automatisierter Blutbilder/Evaluation of computer assisted digital microscopes for evaluation of blood films
  19. Evaluation of computer-assisted digital microscopes for evaluation of blood films1)
https://doi.org/10.1515/labmed-2013-0030
Keywords for this article
automated microscopy; blood film; digital morphology; five-part differential
Creative Commons
BY-NC-ND 3.0

Articles in the same Issue

  1. Masthead
  2. Masthead
  3. Editorial/Leitartikel
  4. Editorial/Leitartikel
  5. Molekulargenetische und zytogenetische Diagnostik/Molecular-Genetic and Cytogenetic Diagnostics
  6. Zellfreie fetale DNA im mütterlichen Blut: neue Möglichkeiten in der pränatalen Diagnostik/Cell free fetal DNA in maternal blood: new possibilities in prenatal diagnosis
  7. Cell-free fetal DNA in maternal blood: new possibilities in prenatal diagnostics1)
  8. Non-invasive prenatal testing for whole chromosome abnormalities/Nicht invasives pränatales Testen auf Chromosomenstörungen
  9. Non-invasive prenatal testing for fetal aneuploidy by massively parallel DNA sequencing of maternal plasma: the future has arrived today/Nicht-invasiver Pränatal-Test auf fetale Aneuploidie mittels massiv-paralleler DNA-Sequenzanalyse im mütterlichen Plasma: in der Zukunft angekommen
  10. Molekulare Karyotypisierung in der Routinediagnostik – Rückblick und Ausblick/Molecular karyotyping in routine diagnostics – a view back and forth
  11. Molecular karyotyping in routine diagnostics – a view back and forth1)
  12. Klinische Chemie und Stoffwechsel/Clinical Chemistry and Metabolism
  13. Vitamin D und kardiovaskuläre Erkrankungen: Standortbestimmung 2012/Vitamin D and cardiovascular diseases: where do we stand in 2012?
  14. Editorial/Leitartikel
  15. 100 Jahre Pappenheimfärbung
  16. Hämatologie/Hematolgoy
  17. Pappenheim-Färbung: Beschreibung einer hämatologischen Standardfärbung – Geschichte, Chemie, Durchführung, Artefakte und Problemlösungen/Pappenheim Stain: Description of a hematological standard stain – history, chemistry, procedure, artifacts and problem solutions
  18. Evaluation von mikroskopischen Systemen zur Erstellung automatisierter Blutbilder/Evaluation of computer assisted digital microscopes for evaluation of blood films
  19. Evaluation of computer-assisted digital microscopes for evaluation of blood films1)
Sign up now to receive a 20% welcome discount
Institutional Access
How does access work?
Have an idea on how to improve our website?
Please write us.
© 2025 De Gruyter Brill
Downloaded on 13.9.2025 from https://www.degruyterbrill.com/document/doi/10.1515/labmed-2013-0030/html
Scroll to top button