Performance evaluation and relevance of the CellaVision™ DM96 system in routine analysis and in patients with malignant hematological diseases

The CellaVision™ DM96 is an automated image analysis system dedicated to locating and preclassifying the various types of white blood cells in peripheral blood smears. The system also partially characterizes of the red blood cell morphology and is able to perform platelet counts. We routinely analyzed the blood samples from 440 patients with quantitative and/or qualitative abnormalities detected by the XE-2100 Sysmex™. Only 2.6% of cells are not identified by DM96™. After classification of the unidentified cells very good correlation coefficients are observed between DM96™ and manual microscopy for most hematological parameters and accuracy is judged excellent up to 98%. For most common parameters, false positive and false negative ratios are also very good. Whatever the pathology and the number of blasts on smear, all patients were positive for blast detection on DM96™. The system is a useful tool for assisting in the diagnosis and classification of most acute or chronic leukemia. Automatic cell location and preclassification, along with unique cell views on the computer screen, could reduce the time spent performing differentials and make real-time collaboration between colleagues a natural part of the classification process. The workstation also provides an ergonomically correct and relaxed working environment. We suggest its use in routine analysis; the system could be very helpful for the accurate morphological diagnosis of samples from patients with malignant hematological disease.


S U M M A R Y
The CellaVision TM DM96 is an automated image analysis system dedicated to locating and preclassifying the various types of white blood cells in peripheral blood smears. The system also partially characterizes of the red blood cell morphology and is able to perform platelet counts. We routinely analyzed the blood samples from 440 patients with quantitative and/or qualitative abnormalities detected by the XE-2100 Sysmex TM . Only 2.6% of cells are not identified by DM96 TM . After classification of the unidentified cells very good correlation coefficients are observed between DM96 TM and manual microscopy for most hematological parameters and accuracy is judged excellent up to 98%. For most common parameters, false positive and false negative ratios are also very good. Whatever the pathology and the number of blasts on smear, all patients were positive for blast detection on DM96 TM . The system is a useful tool for assisting in the diagnosis and classification of most acute or chronic leukemia. Automatic cell location and preclassification, along with unique cell views on the computer screen, could reduce the time spent performing differentials and make real-time collaboration between colleagues a natural part of the classification process. The workstation also provides an ergonomically correct and relaxed working environment. We suggest its use in routine analysis; the system could be very helpful for the accurate morphological diagnosis of samples from patients with malignant hematological disease. quality stained blood smear preparation for the accurate assessment of cellular morphology. Despite the significant improvements during the last years in hematology analyzers, no significant progress has been made in terms of automatic examination of peripheral blood cells. Irrespective of the analyzer, approximately 15% of the blood samples require manual microscopic observation either because of biological rules or analyzer flags. The relative number of samples to be reviewed will probably not decrease in years to come. Smear examinations are time consuming and require well-trained medical technologists and biologists.
Microscopy automation should be available in hematology laboratories. The decrease of cytology proficiency in the daily practice, the need for development of new innovative techniques in hematology laboratories in the face of limited human resources, and finally, the increase and the complexity of pathologies attributable to population aging create a need for automation of the cytology platform in all laboratories. In this context, we had the opportunity to evaluate the CellaVision TM DM96 automated microscope (CellaVision AB, Ideon Research Park, 70 Lund,Sweden) in the hematology laboratory at Caen University Hospital. The academic hospital has 1750 active beds and the laboratory performs 500-600 Complete Blood Count with Differential (CBC-DIFF) per day with XE-2100 Sysmex TM analyzers (Sysmex Corporation 1-S-1, Wakinohama Kaygandori, Cho-Ku, Kobe 651-0073, Japan). Adult and pediatric hematology account for 10% of the demands, oncology represents 15% and surgical and intensive cares about 20%. We evaluated CellaVision TM DM96 and discuss how such a device could be integrated into the daily routine and the performance of DM96 TM in the diagnosis and monitoring of patients with malignant hematological diseases.

M A T E R I A L S A N D M E T H O D S The automated microscope DM 96
CellaVision TM DM96 is an automated device for the differential counting of white blood cells (WBCs) and characterization of red blood cells (RBCs). It consists of a slide feeder unit, a microscope with three objectives (·10, ·50, and ·100), a camera and a computer system containing the acquisition and classification software CellaVision TM blood differential software ( Figure 1). A slide autoloader facilitates the automatic analysis of up to 96 smears with continuous loading access. The number of WBC to be analyzed is user definable from 100 up to 400. To perform a differential count, a thin film of blood is wedged on a glass slide (a blood smear) from a peripheral blood sample and stained according to the May-Grunwald Giemsa protocol. The analyzer performs the acquisition and preclassification of cells and the operator subsequently verifies and modifies, if necessary, the suggested classification of each cell ( Figure 2). The operator can also introduce additional observations and comments when needed. For this reason, persons specially trained in the use of this instrument and skilled in the recognition of cells can operate the DM96 TM . The system makes the following WBC classifications: band neutrophils, segmented neutrophils, eosinophils, basophils, monocytes, lymphocytes, promyelocytes, myelocytes, metamyelocytes, blast cells, variant form lymphocytes and plasma cells. The system also preclassifies non-WBC into the following classifications: erythroblasts, giant thrombocytes, thrombocytes aggregation, smudge cells and artifacts. 'Unidentified' is a class of cells and objects that the system cannot identify. The system has four flag levels for the following RBC morphological characteristics: polychromasia, hypochromasia, anisocytosis, microcytosis, macrocytosis, and poikilocytosis. Besides the WBCs mentioned above, the operator or 'user' can reclassify cells into the following classes afterwards: immature eosinophils, immature basophils, promonocytes, prolymphocytes, large granular lymphocytes, hairy cells, Sezary cells, others, megacaryocyte, not classed, and 15 user-defined classes. The operator can also add the following characteristics for RBCs: schizocytosis, helmet cells, sickle cells, spherocytosis, elliptocytosis, ovalocytosis, teardrop cells, stomatocytosis, acantocytosis, and echinocytosis, Howell-Jolly bodies, Pappenheimer bodies, basophilic stippling, parasites, and 10 other definable characteristics.

Smears and stains
Slides analyzed by the DM96 TM were prepared with SP-100 SYSMEX TM from venous blood sample collected in EDTA-type anticoagulant and previously analyzed with XE-2100 SYSMEX TM . Staining program and reagents were as follows: May Grunwald (MG) and Giemsa (Biolyon, France), MG pure time: 2.5 min, MG dilute time: 3 min, Giemsa time: 7 min, rinse 0 min and drying time 5 min.

Patients
Four hundred and forty nonselected patients processed with the XE-2100 were analyzed by medical technologists experienced using both conventional microscopy method and DM96 TM . All these samples were abnormal according to routine laboratory criteria and hence, justified a manual smear review (quantitative abnormality, qualitative flag from XE-2100 TM , malignant hematological disease). Under the microscope, 100 leucocytes were observed for establishing the control differential and a mean of 110 leucocytes were required for DM96 TM .
To analyze the performance of automated microscopy with DM96 TM and measure its impact on laboratory organization and workflow, we studied its ability to correctly identify blood cells and accuracy compared with manual method and/or XE-2100 TM . Finally, we analyzed the sensitivity for detection of pathological cells in case of hematological disease.

Efficiency of cell recognition
We analyzed the accuracy in classifying normal and abnormal cells for routine parameters by DM96 TM (neutrophils, eosinophils, basophils, lymphocytes, monocytes, immature granulocytes, and erythroblasts) out of 62904 cells [including Nucleated Red Blood Cells (NRBCs) and smudge cells] issued from the 440 patients analyzed. Efficiency of recognition has been calculated for each cell category. Unidentified cells from DM96 TM have been studied as well and their influence on the above-mentioned result calculated.

Comparison with manual method
We did not test normal blood samples in this study as this had already been performed and the results showed DM96 TM to be reliable and accurate (Ceelie, Dinkelaar & van Gelder, 2007). The results of 356 patients with no hematological disease but laboratory flagging criteria obtained on DM96 TM were compared after medical technologist reclassification with the manual differentials performed by the same user and to XE-2100 TM . Correlation between DM96 TM and both manual count and XE-2100 TM result was established for neutrophils, eosinophils, basophils, lymphocytes, monocytes, erythroblasts, and immature granulocytes (including metamyelocytes, myelocytes, and promyelocytes). In case of disagreement, a clinician reanalyzed both the slide and the validation issued from DM96 TM .

Malignant hematological diseases
We focused then on 84 patients with malignant hematological disorders from various types. The classification of these 84 patients was made according to the WHO criteria (Harris et al.,1999) and is described in table 1. Blast recognition and quantification by DM96 TM was studied in 34 patients, acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML) or chronic myeloproliferative disorders/myelodysplastic syndromes (CMPD/MDS). Three patients were excluded for the analysis in the absence of blasts cells in the peripheral blood. All these three patients had myelodysplasic syndromes (MDS). For all other patients, B-cell chronic lymphocytic leukemia (CLL), other B-cell chronic lymphoproliferative disorders (B-CLPD), we focused on capacity of DM96 TM to efficiently recognize mature cells and provide images permitting an easy and reliable morphological classification.

Statistical analysis
Statistical analysis was performed using Microsoft Ò Excel software. For correlation analysis, we used twotailed paired t-tests to evaluate differences between the percentage of blast cells detected by DM96 TM and manual microscope in patients with blast cells in the peripheral blood. Clinical sensitivity and specificity of the CellaVision DM96 TM were defined as its ability to obtain positive and negative results concordant with medical technologist before and after classification of unidentified cells by DM96 TM .

R E S U L T S Accuracy of cell recognition
Only 2.6% of cells are not identified (especially NRBCs and immature granulocytes) leading the global efficiency of DM96 TM to 95% of direct correct identification. In total, when reclassifying unidentified cells by medical technologist, accuracy is judged excellent up to 98%. For most common parameters, false positive and false negative ratio are very good (Table 2).

Comparison with manual method
Correlation for DM96 TM results with the manual method and/or XE-2100 TM is excellent for neutrophils, lymphocytes, eosinophils, and acceptable for immature granulocytes, erythroblasts, and for basophils. The correlation observed for monocytes was not as good as expected; both results (DM96 TM and optical manual count) were usually lower than the automatic count but this was not clinically relevant.

Patients with blasts on smear
Whatever the pathology (AML, ALL, and CMPD/ MDS) and the number of blasts on smear, all 34 patients were positive for blast detection on DM96 TM . Additionally, it appears very easy to distinguish myeloid blasts from lymphoid blasts. Despite DM96 TM underestimates the number of blasts cells (especially in ALL where a huge number of them is misclassified as lymphocytes), after manual validation, the correlation with microscope appears very good ( Figure 3). In patients with chronic myeloproliferative disorders or myelodysplastic syndromes, both the quantitative and qualitative analysis of immature granulocytes was comparable with the one observed for routine patients. Basophiles were clearly identified even when they show abnormal aspects. Blast count for these patients appeared to be reliable.

B-cell chronic lymphoproliferative disorders
DM96 TM classifies CLL cells without problem but often provides a wide count of 'smudge cells' arising from the smear method. The recommended procedure in this case is still to use the lymphocyte count from the analyzer as the most reliable result. Concerning other lymphoid pathologies summarized in

D I S C U S S I O N
The DM96 TM has already proven to be reliable for normal patient samples. In this study, we became aware of the capabilities and limitations of the automated microscope DM96 TM for analyzing patient samples with quantitative and qualitative abnormalities detected by XE-2100 TM . From a biological point of view, we as others (Swolin et al., 2003;Kratz et al., 2005;Roumier et al., 2005;Contis & Williams, 2006, 2006 demonstrated that the results obtained from DM96 TM correlated well to those obtained by manual counting of all patient samples, suggesting that DM96 TM may be useful for the analysis of the great majority of parameters tested. About monocytes, heterogeneous distribution on slides is a well-known problem arising directly from the smear method and consequently, both the choice of the observation area (which can be different for DM96 TM and manual microscopy) and the number of cells analyzed contribute to the difference in monocyte counts. Modern characterization of acute malignant hematological disease is a multidisciplinary process. Initially, it requires the integration of clinical, morphological and cytochemical information. A correct and rapid hematological evaluation is necessary to follow-up with the appropriate laboratory tests, specifically immunophenotyping, metaphase cytogenetics and molecular studies. However, the interpretation of morphological and cytochemical stains remains central to the diagnosis and classification of AML. The identification of multilineage dysplasia is entirely dependent on light microscopic assessment of the leukemia cells. Despite the advances in diagnostic technologies, the maintenance and improvement of morphological skills still remain essential requirements in the diagnosis of AML. In case of AML, DM96 TM is able to detect blast cells and to identify myeloid blast cells and maturing cells. In addition, it facilitates the evaluation and quantification of dysplasia on a high number of myeloid cells preselected by DM96 TM , which are then classified and eventually properly reclassified by the biologist. In cases of B-ALL and more generally in cases of myeloperoxydase (MPO) negative blast cells, immunophenotyping is always still required for the initial diagnosis. DM96 is also able to detect blast cells but is unable to classify blast cells as lymphoblasts. In this context, it makes sense to rely upon conventional microscopy. If the sample quality is poor, it will be also necessary to survey the entire smear in manual mode. For post-treatment monitoring of patients with malignant hematological disease, DM96 TM represents a good tool for the detection of abnormal cells but is not appropriate for quantifying blast cells in the peripheral blood of patients with B-ALL. In cases of B-CLPD and when operated by an experienced cytologist, DM96 TM is helpful for identifying the disease and especially the lymphoid abnormalities. As an example, we rapidly identified binucleated lymphocytes characteristic of polyclonal lymphocytosis with binucleated lymphocytes (Mossafa et al. 1999), hairy cells in patients with HCL and atypical lymphocytes in patients with B-CLL. An overview of all lymphoid cells is of great interest in lymphocyte analysis.
Finally, DM96 proved to be fully comparable with the manual method in a test using control patient blood smears and in daily practice applicable to >90% of the leucocytes review.
The validation and screening of abnormal smears is one of the core competencies of the technical staff, which is under the supervision of a biologist. For such a routine process, a significant timesaving could be realized by implementing such a reliable automatic system. Therefore, these observations should provide food for thought when considering modalities for improving the efficiency of a hematology laboratory.
The routine introduction of DM96 TM will probably have a great impact on the logistics and organization of both specialized and general hematology laboratories. Depending on the validation requirements and guidelines in each country, all the smears performed by SP-100 TM can be passed onto the DM96 TM in a continuous mode. If, after the unidentified cells have been identified, classified, confirmed and validated, there are no blasts present, the validation could be carried out by the DM96 TM , except when particular difficulties are encountered. For the technical staff, the installation of DM96 TM would have many consequences: the reduction of technical staff time at the microscope while simultaneously increasing the efficiency of the workflow, the elimination of medical technologists facing a difficult diagnosis alone, improved ergonomics of the workstation (elbows, eyes, and back), reduction of the filing of the blades and finally, optimization of time and quality. The introduction of DM96 TM could also optimize the time of the biological staff and improve the proficiency of morphological expertise. In addition, the easy and clear presentation of all patient samples on a computer screen will help ensure the quality of follow-up care in cancer patients. The images can be transmitted to other experts for consultation and confirmation and will facilitate validation of clinical protocols. We also can hope for shorter response times, a reduction in errors, an improvement in continued and advanced education, possibly a redeployment of human resources and a significant cost reduction for a hemogram.