Can automated blood film analysis replace the manual differential? An evaluation of the CellaVision DM96 automated image analysis system

White blood cell differential count of maturation stages in bone marrow smear using dual-stage convolutional neural networks

The white blood cell differential count of the bone marrow provides information concerning the distribution of immature and mature cells within maturation stages. The results of such examinations are important for the diagnosis of various diseases and for follow-up care after chemotherapy. However, manual, labor-intensive methods to determine the differential count lead to inter- and intra-variations among the results obtained by hematologists. Therefore, an automated system to conduct the white blood cell differential count is highly desirable, but several difficulties hinder progress. There are variations in the white blood cells of each maturation stage, small inter-class differences within each stage, and variations in images because of the different acquisition and staining processes. Moreover, a large number of classes need to be classified for bone marrow smear analysis, and the high density of touching cells in bone marrow smears renders difficult the segmentation of single cells, which is crucial to traditional image processing and machine learning. Few studies have attempted to discriminate bone marrow cells, and even these have either discriminated only a few classes or yielded insufficient performance. In this study, we propose an automated white blood cell differential counting system from bone marrow smear images using a dual-stage convolutional neural network (CNN). A total of 2,174 patch images were collected for training and testing. The dual-stage CNN classified images into 10 classes of the myeloid and erythroid maturation series, and achieved an accuracy of 97.06%, a precision of 97.13%, a recall of 97.06%, and an F-1 score of 97.1%. The proposed method not only showed high classification performance, but also successfully classified raw images without single cell segmentation and manual feature extraction by implementing CNN. Moreover, it demonstrated rotation and location invariance. These results highlight the promise of the proposed method as an automated white blood cell differential count system.

Performance of automated digital cell imaging analyzer Sysmex DI-60

BACKGROUND

The Sysmex DI-60 system (DI-60, Sysmex, Kobe, Japan) is a new automated digital cell imaging analyzer. We explored the performance of DI-60 in comparison with Sysmex XN analyzer (XN, Sysmex) and manual count.

METHODS

In a total of 276 samples (176 abnormal and 100 normal samples), white blood cell (WBC) differentials, red blood cell (RBC) classification and platelet (PLT) estimation by DI-60 were compared with the results by XN and/or manual count. RBC morphology between pre-classification and verification was compared according to the ICSH grading criteria. The manual count was performed according to the Clinical and Laboratory Standards Institute guidelines (H20-A2).

RESULTS

The overall concordance between DI-60 and manual count for WBCs was 86.0%. The agreement between DI-60 pre-classification and verification was excellent (weighted κ=0.963) for WBC five-part differentials. The correlation with manual count was very strong for neutrophils (r=0.955), lymphocytes (r=0.871), immature granulocytes (r=0.820), and blasts (r=0.879). RBC grading showed notable differences between DI-60 and manual counting on the basis of the ICSH grading criteria. Platelet count by DI-60 highly correlated with that by XN (r=0.945). However, DI-60 underestimated platelet counts in samples with marked thrombocytosis.

CONCLUSIONS

The performance of DI-60 for WBC differential, RBC classification, and platelet estimation seems to be acceptable even in abnormal samples with improvement after verification. DI-60 would help optimize the workflow in hematology laboratory with reduced manual workload.

Very Deep Convolutional Neural Networks for Morphologic Classification of Erythrocytes

BACKGROUND

Morphologic profiling of the erythrocyte population is a widely used and clinically valuable diagnostic modality, but one that relies on a slow manual process associated with significant labor cost and limited reproducibility. Automated profiling of erythrocytes from digital images by capable machine learning approaches would augment the throughput and value of morphologic analysis. To this end, we sought to evaluate the performance of leading implementation strategies for convolutional neural networks (CNNs) when applied to classification of erythrocytes based on morphology.

METHODS

Erythrocytes were manually classified into 1 of 10 classes using a custom-developed Web application. Using recent literature to guide architectural considerations for neural network design, we implemented a "very deep" CNN, consisting of >150 layers, with dense shortcut connections.

RESULTS

The final database comprised 3737 labeled cells. Ensemble model predictions on unseen data demonstrated a harmonic mean of recall and precision metrics of 92.70% and 89.39%, respectively. Of the 748 cells in the test set, 23 misclassification errors were made, with a correct classification frequency of 90.60%, represented as a harmonic mean across the 10 morphologic classes.

CONCLUSIONS

These findings indicate that erythrocyte morphology profiles could be measured with a high degree of accuracy with "very deep" CNNs. Further, these data support future efforts to expand classes and optimize practical performance in a clinical environment as a prelude to full implementation as a clinical tool.

Spectral-spatial feature-based neural network method for acute lymphoblastic leukemia cell identification via microscopic hyperspectral imaging technology

Microscopic examination is one of the most common methods for acute lymphoblastic leukemia (ALL) diagnosis. Most traditional methods of automized blood cell identification are based on RGB color or gray images captured by light microscopes. This paper presents an identification method combining both spectral and spatial features to identify lymphoblasts from lymphocytes in hyperspectral images. Normalization and encoding method is applied for spectral feature extraction and the support vector machine-recursive feature elimination (SVM-RFE) algorithm is presented for spatial feature determination. A marker-based learning vector quantization (MLVQ) neural network is proposed to perform identification with the integrated features. Experimental results show that this algorithm yields identification accuracy, sensitivity, and specificity of 92.9%, 93.3%, and 92.5%, respectively. Hyperspectral microscopic blood imaging combined with neural network identification technique has the potential to provide a feasible tool for ALL pre-diagnosis.

Characterization and automatic screening of reactive and abnormal neoplastic B lymphoid cells from peripheral blood

INTRODUCTION

The objective was to advance in the automatic, image-based, characterization and recognition of a heterogeneous set of lymphoid cells from peripheral blood, including normal, reactive, and five groups of abnormal lymphocytes: hairy cells, mantle cells, follicular lymphoma, chronic lymphocytic leukemia, and prolymphocytes.

METHODS

A number of 4389 images from 105 patients were selected by pathologists, based on morphologic visual appearance, from patients whose diagnosis was confirmed by all the remaining complementary tests. Besides geometry, new color and texture features were extracted using six alternative color spaces to obtain rich information to characterize the cell groups. The recognition system was designed using support vector machines trained with the whole image set.

RESULTS

In the experimental tests, individual sets of images from 21 new patients were analyzed by the trained recognition system and compared with the true diagnosis. An overall recognition accuracy of 97.67% was achieved when the cell screening was performed into three groups: normal lymphocytes, abnormal lymphoid cells, and reactive lymphocytes. The accuracy of the whole experimental study was 91.23% when considering the further discrimination of the abnormal lymphoid cells into the specific five groups.

CONCLUSION

The excellent automatic screening of the three groups of normal, reactive, and abnormal lymphocytes is useful as it discriminates between malignancy and not malignancy. The discrimination of the five groups of abnormal lymphoid cells is encouraging toward the idea that the system could be an automated image-based screening method to identify blood involvement by a variety of B lymphomas.

Impact of Integrating Rumke Statistics to Assist with Choosing Between Automated Hematology Analyzer Differentials vs Manual Differentials

BACKGROUND

To optimize precision of nucleated blood cell counting, the clinical laboratory scientist should post the automated differential rather than the manual differential if the former is within the 95% CI of the latter, as determined by the "Rumke statistic." The objective of this study was to determine the potential impact of real-time, computer-assisted use of Rumke statistics for more judicious use of the automated vs digitally imaged, manual differential.

METHODS

Complete blood counts with automated differentials produced by a XE5000™ hematology analyzer (Sysmex) were compared with both the DM96 (CellaVision™ AB) preclassification differentials and the posted reclassified manual differentials, using the Rumke 95% CIs as calculated using the Clopper-Pearson method.

RESULTS

A total of 44.7% of manual differentials had no statistical or clinical justification over the automated differential. In addition, 31.1% of manual differentials had statistical discrepancies between the instrument absolute neutrophil count (IANC) and manual absolute neutrophil count (ANC). Nineteen of these IANC/manual ANC discrepant samples had ANCs below 1500/μL, a decision level for cancer treatment. Holding the IANC when it is less than 2000/μL until after manual smear review would have prevented the posting of any IANC vs manual ANC discrepant results at the 1500/μL ANC decision threshold.

CONCLUSIONS

A real-time operator alert concerning the statistical identity of imaging device differentials vs automated differentials could have reduced manual differentials by nearly 45%. Not posting unnecessary manual differentials for the cases with IANC/manual ANC discrepancies would have likely reduced clinical error/confusion.

Feature Analysis and Automatic Identification of Leukemic Lineage Blast Cells and Reactive Lymphoid Cells from Peripheral Blood Cell Images

BACKGROUND

Automated peripheral blood (PB) image analyzers usually underestimate the total number of blast cells, mixing them up with reactive or normal lymphocytes. Therefore, they are not able to discriminate between myeloid or lymphoid blast cell lineages. The objective of the proposed work is to achieve automatic discrimination of reactive lymphoid cells (RLC), lymphoid and myeloid blast cells and to obtain their morphologic patterns through feature analysis.

METHODS

In the training stage, a set of 696 blood cell images was selected in 32 patients (myeloid acute leukemia, lymphoid precursor neoplasms and viral or other infections). For classification, we used support vector machines, testing different combinations of feature categories and feature selection techniques. Further, a validation was implemented using the selected features over 220 images from 15 new patients (five corresponding to each category).

RESULTS

Best discrimination accuracy in the training was obtained with feature selection from the whole feature set (90.1%). We selected 60 features, showing significant differences (P < 0.001) in the mean values of the different cell groups. Nucleus-cytoplasm ratio was the most important feature for the cell classification, and color-texture features from the cytoplasm were also important. In the validation stage, the overall classification accuracy and the true-positive rates for RLC, myeloid and lymphoid blast cells were 80%, 85%, 82% and 74%, respectively.

CONCLUSION

The methodology appears to be able to recognize reactive lymphocytes well, especially between reactive lymphocytes and lymphoblasts.

Evaluation of the accuracy of the CellaVision™ DM96 in a high HIV-prevalence population in South Africa

INTRODUCTION

The CellaVision™ DM96 (DM96) is a digital microscopy system which performs well in developed countries. However, to date it has not been evaluated in Africa, where the pathology spectrum encountered is very different. In particular, its utility in a setting with high HIV prevalence has not been assessed, which is of interest because of the morphological aberrations often seen in HIV-positive patients.

OBJECTIVES

This study aimed to evaluate the accuracy of the DM96 in a South African laboratory, with emphasis on its performance in samples collected from HIV-positive patients.

METHODS

A total of 149 samples submitted for a routine differential white cell count in 2012 and 2013 at the Chris Hani Baragwanath Academic Hospital in Johannesburg, South Africa were included, of which 79 (53.0%) were collected from HIV-positive patients. Results of DM96 analysis pre- and post-classification were compared with a manual differential white cell count and the impact of HIV infection and other variables of interest were assessed.

RESULTS

Pre- and post-classification accuracies were similar to those reported in developed countries. Reclassification was required in 16% of cells, with particularly high misclassification rates for eosinophils (31.7%), blasts (33.7%) and basophils (93.5%). Multivariate analysis revealed a significant relationship between the number of misclassified cells and both the white cell count (p = 0.035) and the presence of malignant cells in the blood (p = 0.049), but not with any other variables analysed, including HIV status.

CONCLUSION

The DM96 exhibited acceptable accuracy in this South African laboratory, which was not impacted by HIV infection. However, as it does not eliminate the need for experienced morphologists, its cost may be unjustifiable in a resource-constrained setting.

Performance of the CellaVision(®) DM96 system for detecting red blood cell morphologic abnormalities

Background:

Red blood cell (RBC) analysis is a key feature in the evaluation of hematological disorders. The gold standard light microscopy technique has high sensitivity, but is a relativity time-consuming and labor intensive procedure. This study tested the sensitivity and specificity of gold standard light microscopy manual differential to the CellaVision^® DM96 (CCS; CellaVision, Lund, Sweden) automated image analysis system, which takes digital images of samples at high magnification and compares these images with an artificial neural network based on a database of cells and preclassified according to RBC morphology.

Methods:

In this study, 212 abnormal peripheral blood smears within the Calgary Laboratory Services network of hospital laboratories were selected and assessed for 15 different RBC morphologic abnormalities by manual microscopy. The same samples were reassessed as a manual addition from the instrument screen using the CellaVision^® DM96 system with 8 microscope high power fields (×100 objective and a 22 mm ocular). The results of the investigation were then used to calculate the sensitivity and specificity of the CellaVision^® DM96 system in reference to light microscopy.

Results:

The sensitivity ranged from a low of 33% (RBC agglutination) to a high of 100% (sickle cells, stomatocytes). The remainder of the RBC abnormalities tested somewhere between these two extremes. The specificity ranged from 84% (schistocytes) to 99.5% (sickle cells, stomatocytes).

Conclusions:

Our results showed generally high specificities but variable sensitivities for RBC morphologic abnormalities.

Automatic recognition of atypical lymphoid cells from peripheral blood by digital image analysis

OBJECTIVES

The objective was the development of a method for the automatic recognition of different types of atypical lymphoid cells.

METHODS

In the method development, a training set (TS) of 1,500 lymphoid cell images from peripheral blood was used. To segment the images, we used clustering of color components and watershed transformation. In total, 113 features were extracted for lymphocyte recognition by linear discriminant analysis (LDA) with a 10-fold cross-validation over the TS. Then, a new validation set (VS) of 150 images was used, performing two steps: (1) tuning the LDA classifier using the TS and (2) classifying the VS in the different lymphoid cell types.

RESULTS

The segmentation algorithm was very effective in separating the cytoplasm, nucleus, and peripheral zone around the cell. From them, descriptive features were extracted and used to recognize the different lymphoid cells. The accuracy for the classification in the TS was 98.07%. The precision, sensitivity, and specificity values were above 99.7%, 97.5%, and 98.6%, respectively. The accuracy of the classification in the VS was 85.33%.

CONCLUSIONS

The method reaches a high precision in the recognition of five different types of lymphoid cells and could allow for the design of a diagnosis support tool in the future.

Development and validation of an app-based cell counter for use in the clinical laboratory setting

Introduction:

For decades cellular differentials have been generated exclusively on analog tabletop cell counters. With the advent of tablet computers, digital cell counters – in the form of mobile applications (“apps”) – now represent an alternative to analog devices. However, app-based counters have not been widely adopted by clinical laboratories, perhaps owing to a presumed decrease in count accuracy related to the lack of tactile feedback inherent in a touchscreen interface. We herein provide the first systematic evidence that digital cell counters function similarly to standard tabletop units.

Methods:

We developed an app-based cell counter optimized for use in the clinical laboratory setting. Paired counts of 188 peripheral blood smears and 62 bone marrow aspirate smears were performed using our app-based counter and a standard analog device. Differences between paired data sets were analyzed using the correlation coefficient, Student's t-test for paired samples and Bland–Altman plots.

Results:

All counts showed excellent agreement across all users and touch screen devices. With the exception of peripheral blood basophils (r = 0.684), differentials generated for the measured cell categories within the paired data sets were highly correlated (all r ≥ 0.899). Results of paired t-tests did not reach statistical significance for any cell type (all P > 0.05), and Bland–Altman plots showed a narrow spread of the difference about the mean without evidence of significant outliers.

Conclusions:

Our analysis suggests that no systematic differences exist between cellular differentials obtained via app-based or tabletop counters and that agreement between these two methods is excellent.

Flow cytometric white blood cell differential using CytoDiff is excellent for counting blasts

BACKGROUND

The usefulness of the CytoDiff flow cytometric system (Beckman Coulter, USA) has been studied in various conditions, but its performance including rapidity in detecting and counting blasts, the most significant abnormal cells in the peripheral blood, has not been well evaluated. The objective of this study was to evaluate the performance of the CytoDiff differential counting method in challenging samples with blasts.

METHODS

In total, 815 blood samples were analyzed. Samples flagged as "blasts" or "variant lymphocytes" and showing <10% blasts by manual counts were included. In total, 322 samples showed blasts on manual counts, ranging from 0.5% to 99%. The CytoDiff method was performed by flow cytometry (FC500; Beckman Coulter, USA) with a pre-mixed CytoDiff reagent and analyzing software (CytoDiff CXP 2.0; Beckman Coulter).

RESULTS

The average time required to analyze 20 samples was approximately 60 min for manual counts, and the hands-on time for the CytoDiff method was 15 min. The correlation between the CytoDiff and manual counts was good (r>0.8) for neutrophils and lymphocytes but poor (r<0.8) for other cells. When the cutoff value of the CytoDiff blast count was set at 1%, the sensitivity was 94.4% (95% CI; 91.2-96.6) and specificity was 91.9% (95% CI; 89.0-94.1). The positive predictive value was 88.4% (95% CI; 84.4-91.5) (304/344 cases) and negative predictive value was 96.2% (95% CI; 93.9-97.7) (453/471 cases). The CytoDiff blast counts correlated well to the manual counts (r=0.9223).

CONCLUSIONS

The CytoDiff method is a specific, sensitive, and rapid method for counting blasts. A cutoff value of 1% of at least 1 type of blast is recommended for positive CytoDiff blast counts.

Accuracy of the CellaVision DM96 platform for reticulocyte counting

CONTEXT

Many hematology laboratories have adopted semi-automated digital platforms for routine use and the evidence supporting their use is increasing.

AIMS

The CellaVision platforms are among the most thoroughly studied digital hematology platforms; we wished to determine the accuracy of CellaVision for reticulocyte counting.

DESIGN MATERIALS AND METHODS

We compared reticulocyte counts performed manually, using the Beckman Coulter LH750 automated analyzer and with the CellaVision DM96 platform. We analyzed the results for pair-wise correlation and bias, and precision.

STATISTICAL ANALYSES USED

Analyses were performed using Statistical Package for the Social Sciences software (SPSS), including Spearman's rho correlation coefficient, Friedman's two-way Analysis Of Variance (ANOVA) for comparison of distributions; bias was compared by way of mean and standard deviation.

RESULTS

The CellaVision reticulocyte counts correlated most strongly with those of the analyzer (often considered the benchmark test); the reticulocyte count distributions were noted not to be significantly different from each other across all three methods. The mean and standard deviation of bias were lowest in the comparison of CellaVision and LH750 counts.

CONCLUSIONS

Our data provide additional support for the accuracy of digital hematology applications using the CellaVision DM96 platform.

Validation of the Sysmex sp-1000i automated slide preparer-stainer in a clinical laboratory

BACKGROUND

The speed and quality of information have become essential items in the release of laboratory reports. The Sysmex(®)SP1000-I device has been developed to prepare and stain smear slides. However, for a device to be cleared for use in the laboratory routine it must pass through a validation process.

OBJECTIVE

To evaluate the performance and reliability of the Sysmex(®) SP-1000i slide preparer-stainer incorporated into the routine of a hospital laboratory in Porto Alegre.

METHODS

Peripheral blood samples of patients attending the laboratory for ambulatory exams with leukocyte counts between 7000/°L and 12,000/°L were evaluated, independent of gender and age. Two slides were prepared for each sample using the Sysmex(®) SP-1000i equipment; one of the slides was used to perform quality control tests using the CellaVision(®) DM96 device, and the other slide was used to compare pre-classification by the same device and the classification performed by a pharmacist-biochemist.

RESULTS

The results of all the slides used as controls were acceptable according to the quality control test as established by the manufacturer of the device. In the comparison between the automated pre-classification and the classification made by the professional, there was an acceptable variation in the differential counts of leukocytes for 90% of the analyzed slides. Pearson correlation coefficient showed a strong correlation for band neutrophils (r = 0.802; p-value < 0.001), segmented neutrophils (r = 0.963; p-value < 0.001), eosinophils (r = 0.958; p-value < 0.001), lymphocytes (r = 0.985; p-value < 0.001) and atypical lymphocytes (r = 0.866; p-value < 0.001) using both methods. The red blood cell analysis was adequate for all slides analyzed by the equipment and by the professional.

CONCLUSION

The new Sysmex(®)SP1000-i methodology was found to be reliable, fast and safe for the routines of medium and large laboratories, improving the quality of microscopic analysis in complete blood counts.

Platelet count estimation using the CellaVision DM96 system

INTRODUCTION

Rapid and accurate determination of platelet count is an important factor in diagnostic medicine. Traditional microscopic methods are labor intensive with variable results and are highly dependent on the individual training. Recent developments in automated peripheral blood differentials using a computerized system have shown many advantages as a viable alternative. The purpose of this paper was to determine the reliability and accuracy of the CellaVision DM 96 system with regards to platelet counts.

MATERIALS AND METHODS

One hundred twenty seven peripheral blood smears were analyzed for platelet count by manual microscopy, an automated hematology analyzer (Beckman Counter LH 780 or Unicel DXH 800 analyzers) and with the CellaVision DM96 system. Results were compared using the correlations and Bland-Altman plots.

RESULTS

Platelet counts from the DM96 system showed an R(2) of 0.94 when compared to manual platelet estimates and an R(2) of 0.92 when compared to the automated hematology analyzer results. Bland-Altman plots did not show any systematic bias.

Performance of CellaVision DM96 in leukocyte classification

Background:

Leukocyte differentials are an important component of clinical care. Morphologic assessment of peripheral blood smears (PBS) may be required to accurately classify leukocytes. However, manual microscopy is labor intensive. The CellaVision DM96 is an automated system that acquires digital images of leukocytes on PBS, pre-classifies the cell type, and displays them on screen for a Technologist or Pathologist to approve or reclassify. Our study compares the results of the DM96 with manual microscopy.

Methods:

Three hundred and fifty-nine PBS were selected and assessed by manual microscopy with a 200 leukocyte cell count. They were then reassessed using the CellaVision DM96 with a 115 leukocyte cell count including reclassification when necessary. Correlation between the manual microscopy results and the CellaVision DM96 results was calculated for each cell type.

Results:

The correlation coefficients (r²) range from a high of 0.99 for blasts to a low of 0.72 for metamyelocytes.

Conclusions:

The correlation between the CellaVision DM96 and manual microscopy was as good or better than the previously published data. The accuracy of leukocyte classification depended on the cell type, and in general, there was lower correlation for rare cell types. However, the correlation is similar to previous studies on the correlation of manual microscopy with an established reference result. Therefore, the CellaVision DM96 is appropriate for clinical implementation.

Experience with CellaVision DM96 for peripheral blood differentials in a large multi-center academic hospital system

Context and Aims:

Rapid, accurate peripheral blood differentials are essential to maintain standards of patient care. CellaVision DM96 (CellaVision AB, Lund, Sweden) (CV) is an automated digital morphology and informatics system used to locate, pre-classify, store and transmit images of platelets, red and white blood cells to a trained technologist who confirms or edits CV cell classification. We assessed our experience with CV by evaluating sensitivity, specificity, positive predictive value and negative predictive value for CV in three different patient populations.

Materials and Methods:

We analyzed classification accuracy of CV for white blood cells, erythroblasts, platelets and artefacts over six months for three different university hospitals using CV.

Results:

CV classified 211,218 events for the adult cancer center; 51,699 events for the adult general hospital; and 8,009 events for the children's hospital with accuracy of CV being 93%, 87.3% and 95.4% respectively. Sensitivity and positive predictive value were <80% for immature granulocytes (band neutrophil, promyelocyte, myelocyte and metamyelocytes) (differences usually within one stage of maturation). Cell types comprising a lower frequency of the total events, including blasts, showed lower accuracy at some sites.

Conclusions:

The reduced immature granulocyte classification accuracy may be due in part to the subjectivity in classification of these cells, length of experience with the system and individual expertise of the technologist. Cells with low sensitivity and positive predictive value comprised a minority of the cells and should not significantly affect the technologist re-classification time. CV serves as a clinically useful instrument in performance of peripheral blood differentials.

Toward a reference method for leukocyte differential counts in blood: comparison of three flow cytometric candidate methods

A Complete Blood Count performed by an automated hematology analyzer frequently requires a microscopic slide review. Recently, we and others have proposed combinations of monoclonal antibodies for an extended leukocyte differential by flow cytometry. The aim of this study was to compare the performance of these proposals. Ninety-two samples were analyzed at 2 sites to compare the accuracy of three published methods. Reference methods used were i) cell counter for leukocyte count and ii) microscopic review as defined by CSLI H20-A2 for cell subsets. Comparison of flow cytometers from 2 manufacturers (FC500 and CANTO/LSRII) was performed. Published protocols were adapted to three different models of flow cytometer and each provided similar results in leukocyte subset enumeration, although some discrepancies were noted for each protocol in comparison with the reference method. The conclusion is that each protocol carries advantages and disadvantages and there is no clear "winner". This study supports the fact that flow cytometry is a candidate to become a reference method for the leukocyte differential. None of the tested protocols clearly demonstrated superiority and each had demonstrable deficiencies. Additional work to develop a consensual 8 to 10 color panel is concluded to be necessary for a satisfactory reference method.

Evaluation of an Automated Digital Imaging System, Nextslide Digital Review Network, for Examination of Peripheral Blood Smears

Context.—Several automated digital imaging systems have been introduced in recent years to improve turnaround time and proficiency in examining peripheral blood smears in hematology laboratories.

Objective.—To evaluate a new automated digital imaging system, Nextslide Digital Review Network (Nextslide), for examination of peripheral blood smears.

Design.—We evaluated 479 peripheral blood smears, of which 247 (51.6%) were included for comparison of Nextslide and manual white blood cell differential counts and morphology evaluation, 204 (42.6%) were included for comparison of Nextslide and CellaVision (DM96) differential counts, and 28 (5.8%) were neonatal samples examined for enumeration of nucleated red blood cells.

Results.—Results from both method comparisons showed excellent correlation for all major white blood cell classes with correlation coefficients ranging from 0.70 to 0.99. Evaluation of white blood cell, red blood cell, and platelet morphology also showed good correlation among methods. White blood cell preclassification capability in the system was evaluated for rate and accuracy. Leukopenic samples demonstrated markedly decreased review time with Nextslide. Enumeration of nucleated red blood cells showed good correlation among methods.

Conclusions.—Our evaluation of Nextslide shows excellent correlation when compared with conventional manual differentials and CellaVision (DM96) differentials for evaluation of peripheral blood smears.

Absolute neutrophil counts from automated hematology instruments are accurate and precise even at very low levels

Using 106 samples from patients with an absolute neutrophil count (ANC) less than 2.0 × 10(9)/L, two 5-part differential hematology instruments (Sysmex XE-2100, Sysmex, Kobe, Japan, and Advia 2120i, Siemens Healthcare Diagnostics, Deerfield, IL), two 3-part differential hematology instruments (Sysmex K4500, Sysmex, and Advia 60, Siemens Healthcare Diagnostics), and an automated system for examination of microscopic slides (CellaVision DM96, CellaVision, Lund, Sweden) were compared with a flow cytometric (FCM) neutrophil count using monoclonal antibodies for cell classification. The precision and accuracy of the 5-part differential instrument ANC was very good at more than 0.1 × 10(9)/L, although a small systematic difference (10.3%) was found between the 2 instruments. The ANC of the 3-part differential instruments was less reliable, but the WBC count correlated very well with the WBC count from the 5-part differential instruments. Also, the neutrophil count from the CellaVision DM96 compared very well with FCM. When used in the correct laboratory setting, all of the evaluated instruments provide ANCs and WBCs with adequate accuracy and precision.

Canine differential leukocyte counting with the CellaVision DM96Vision, Sysmex XT-2000iV, and Advia 2120 hematology analyzers and a manual method

BACKGROUND

For differential leukocyte counts, automated blood smear evaluation systems have been too slow or inaccurate to replace or supplement the manual differential count. The CellaVision DM96Vision (DM96V), a new instrument, is an automated image analysis system that is rapid and accurate enough to be used for enumerating human leukocytes and may be useful for analysis of canine blood.

OBJECTIVES

The aims of this study were to evaluate the performance of the DM96V in differential counting of canine leukocytes, to compare its performance with that of other methods, and to analyze interoperator variability.

METHODS

Four methods of determining the leukocyte differential count of 108 canine blood samples were compared based on agreement, precision, and errors as well as relative performance. Differential counts were obtained using the DM96V, the manual method, and automated methods performed by the Advia 2120 and Sysmex XT-2000iV.

RESULTS

All leukocyte types were detected by the DM96V and the manual method, and all 4 methods had similar mean and median results in most cases. The automated methods were more precise than either the DM96V or manual method when comparing identification of a single type of leukocyte, especially neutrophils and lymphocytes. However, precision of the automated methods was only fair for monocytes, and the Advia and Sysmex failed to identify basophils. The Advia reported fewer monocytes and eosinophils than did the other methods. Significantly fewer lymphocytes were identified by the manual method than by the Sysmex, Advia, and DM96V. The DM96V occasionally presented duplicate images of the same neutrophils.

CONCLUSIONS

The CellaVision DM96V is a satisfactory system for facilitating canine differential leukocyte counting. The DM96V differential count was more similar to the manual count than to automated counts, which were more precise but had errors and omissions in detecting some types of leukocytes.

Image microarrays (IMA): Digital pathology's missing tool

Introduction:

The increasing availability of whole slide imaging (WSI) data sets (digital slides) from glass slides offers new opportunities for the development of computer-aided diagnostic (CAD) algorithms. With the all-digital pathology workflow that these data sets will enable in the near future, literally millions of digital slides will be generated and stored. Consequently, the field in general and pathologists, specifically, will need tools to help extract actionable information from this new and vast collective repository.

Methods:

To address this limitation, we designed and implemented a tool (dCORE) to enable the systematic capture of image tiles with constrained size and resolution that contain desired histopathologic features.

Results:

In this communication, we describe a user-friendly tool that will enable pathologists to mine digital slides archives to create image microarrays (IMAs). IMAs are to digital slides as tissue microarrays (TMAs) are to cell blocks. Thus, a single digital slide could be transformed into an array of hundreds to thousands of high quality digital images, with each containing key diagnostic morphologies and appropriate controls. Current manual digital image cut-and-paste methods that allow for the creation of a grid of images (such as an IMA) of matching resolutions are tedious.

Conclusion:

The ability to create IMAs representing hundreds to thousands of vetted morphologic features has numerous applications in education, proficiency testing, consensus case review, and research. Lastly, in a manner analogous to the way conventional TMA technology has significantly accelerated in situ studies of tissue specimens use of IMAs has similar potential to significantly accelerate CAD algorithm development.

Improved flagging rates on the Sysmex XE-5000 compared with the XE-2100 reduce the number of manual film reviews and increase laboratory productivity

Hematology analyzers generate suspect flags in the presence of abnormal cells. False-positive rates for flags are high on all analyzers. Sysmex, Kobe, Japan, has developed new software for its XE-5000 with improved algorithms for flagging blast cells, abnormal lymphocytes or lymphoblasts, and atypical lymphocytes. This study evaluated the efficiency of these flags in 1,002 samples. The XE-5000 was compared with the XE-2100 (Sysmex) and microscopic examination of cell morphologic features. On the XE-2100, the blast flag demonstrated 90 false-positives, 13 true-positives, and 3 false-negatives. The values on the XE-5000 were 27 false-positives, 14 true-positives, and 2 false-negatives. The abnormal lymphocyte/lymphoblast flag was assessed with the atypical lymphocyte flag. The XE-2100 showed 114 false-positives, 23 true-positives, and 20 false-negatives, and on the XE-5000, there were 45 false-positives, 22 true-positives, and 21 false-negatives. This more specific flagging reduces the number of films that require manual review.

Leukoflow: multiparameter extended white blood cell differentiation for routine analysis by flow cytometry

Differential white blood cell count (dWBC) is a frequently used diagnostic tool. For most patient samples an automated blood counter produces a five-part differential count. If this dWBC does not meet pre-set criteria, microscopic dWBC is performed. Microscopy is labor intensive and requires sustained training of technicians. Inter-observer variation and statistical variation are significant, due to limited numbers of cells counted. Flow cytometry is a candidate reference method for dWBC. Advantages are immunological definitions and large number of measured cells. Our goal was to replace (part of) the microscopic dWBC by a flow cytometric dWBC, that gives additional information on blasts, myeloid precursors, and lymphocyte subsets. We designed a cocktail of antibodies (CD4, CD14, CD34, CD16, CD56, CD19, CD45, CD138, CD3, and CD71) combined with a gating strategy and flow cytometric protocol for easy identification of leukocyte populations. This assay, called Leukoflow, requires low sample volume, has few manual handling steps, and a potential turn-around-time shorter than 2 h. We determine percentages and absolute concentrations of at least 13 different cell populations. For quantification of normoblasts a second flow cytometric staining was designed. We compared microscopic dWBC with that of the automated blood counter and Leukoflow for normal and abnormal blood samples. Leukoflow results correlate well with the automated blood counter for leukocytes, neutrophils, eosinophils, monocytes, and lymphocytes. Correlation with manual dWBC is lower. Blast counts reported by Leukoflow suffer less from inter-observer variation compared to manual dWBC. In addition to microscopic or cytometric dWBC-techniques T-lymphocytes, CD4-T-lymphocytes, B-lymphocytes, NK-cells, myeloid progenitors, plasma cells, and blasts are determined by Leukoflow. These populations give potential useful clinical information and are subject for future studies focusing on the additional clinical value. Leukoflow is a highly interesting and promising technique for clinical laboratories.