Jurgen A Riedl

Automated morphological analysis of cells in body fluids by the digital microscopy system DM96

Background Differential counting and morphological analysis of nucleated cells in body fluids (eg, cerebrospinal fluid and pleural fluid) are of great diagnostic importance to the clinician. A recent development in this field was the introduction of an application for an automated microscopy system, the DM96 Body Fluid module, enabling the automated analysis of body fluid samples. This computerised system provides an automated morphological analysis of body fluids, including an automated classification of all nucleated cells. Aims To investigate the ability of the digital microscopy system, DM96, to automatically classify cells in different types of body fluids. Methods A total of 177 body fluids (including cerebrospinal fluid, abdominal fluid and continuous ambulant peritoneal dialysis fluid) were analysed on the DM96, and results were compared with the manual microscopy method. Results A study in 177 samples demonstrates an overall preclassification accuracy of 90% in spinal fluid and 83% in other body fluids using the automated system. Correlation coefficients for postclassification as compared with manual review range from 0.92 to 0.99 for spinal fluid sample analyses and from 0.83 to 0.98 for other body fluids. The within-run variation of automated classification is less than 6% for all cell categories (4% excluding macrophages). Conclusion The DM96 has proven to be reliable and efficient, contributing to overall quality improvement in morphological analysis and automated cell classification of peripheral blood and other body-fluid samples.

Classification of several morphological red blood cell abnormalities by DM96 digital imaging

Sir, Manual light microscopy remains the gold standard for morphological analysis of peripheral blood smears. However, this method is labour-intensive and time-consuming and requires highly trained personnel. Digital microscope systems contribute in realizing a more rapid, standardized and efficient morphological analysis of peripheral blood smears [1, 2]. Automatic classification of white blood cells (WBCs) using digital imaging in a peripheral blood smear is already well established for the five main classes (neutrophils, lymphocytes, monocytes, eosinophils and basophils) and blast cells [2–4]. These cell classes can be recognized with a high degree of reliability; less common classes such as promyelocytes, myelocytes and metamyelocytes are also recognized by the system but with less accuracy [2]. The use of digital microscopy systems could also improve morphological analysis of aberrant red blood cells (e.g. schistocytes, teardrop cells), eventually leading to a faster diagnosis and treatment of diseases in which these abnormalities are the hallmark of these diseases. Moreover, digital microscopy will lead to a more standardized way of performing these kinds of analyses. The recent introduction of a novel application module for the digital microscope system DM96 (Advanced RBC application, CellaVision, Lund, Sweden) allows for the automatic detection and classification of morphological abnormalities in red blood cells. This application was developed using an artificial neural network, and it considers 80 features such as size, roundness, size and shape of inner pallor and distribution of notches around the border for the morphological classification of erythrocytes. The system generates an image (Figure 1), which corresponds with an area of eight microscopic fields (1009 objective and a 22 mm ocular). While producing this image, the system also performs a preclassification of the red blood cells (RBCs). The classification results of the morphological abnormality are displayed on the screen and can subsequently be altered or confirmed by the operator. The module has already been validated for the detection and subsequent classification of teardrop cells and schistocytes in peripheral blood smears [5–7]. To date, standardization of red blood cell morphology is still limited [8, 9]. In this study, we continued our ongoing validation of the RBC module by comparing postclassification results (after manual intervention by three morphological experts) from the RBC module on the DM96 to results from manual assessment, despite all still the gold standard. We also compared the RBC module postclassification results to the preclassification results by the software (without manual interference). For this, a cohort of patient samples and normal samples were used. Result Laboratory serves as the reference laboratory for blood cell morphology proficiency testing for the Netherlands and morphology experts from this department are highly experienced, trained and skilled technicians. In addition, the department participates in internal and external quality control procedures on a routine basis. The morphological experts using the RBC module were tested for competence and examined (score of minimal 90% concordance with examinator) before working on this study. No clinical significant discrepancies were present between the morphological experts. A total of 316 peripheral blood smears were used for analysis, including 198 patient samples and 118 normal samples (according to manual assessment). Due to limited availability of positive samples, the morphological abnormalities elliptocytes, ovalocytes, echinocytes, pappenheimer bodies, basophilic stippling and parasites were excluded from this study. The included abnormalities are listed in Table 1. Due to clinical relevance, different guidelines were used for the percentages of abnormal cells. The cut-off values for the various red blood cell abnormalities described in this article are the values currently used in our laboratory. The peripheral blood smears were prepared from venous blood samples collected in EDTA tubes, using the SP-10 (slide maker-stainer, Sysmex, Etten-Leur, the

Automated detection and classification of teardrop cells by a novel RBC module using digital imaging/microscopy

Sir, In a number of diseases, morphological analysis of abnormal red blood cells is of vital importance, contributing to a rapid and correct diagnosis. An example is teardrop cells, which may indicate dyserythropoiesis due to circulation through altered bone marrow sinuses and splenic cords. Teardrop cells in a peripheral blood smear are significant and correlate with several diseases, such as iron deficiency anemia, hemolytic anemia, megaloblastic anemia, and metastatic carcinoma due to bone marrow infiltration and myelofibrosis [1, 2]. Myelofibrosis is characterized by a progressive anemia, pancytopenia, and thrombocytopenia. Typically present are splenomegaly and/or hepatomegaly due to extramedullary erythropoiesis, leukoerythroblastosis caused by the lack of space in the bone marrow due to fibrosis, and the presence of myeloid precursors and teardrop cells in peripheral blood. Nowadays, the grading of abnormal red blood cells is semi-quantitative, usually indicated by a score ranging from 1 to 3 + correlating with the percentage of abnormal cells present. A consistent and standardized international grading system for abnormal red blood cells is still lacking, resulting in scoring percentages that differ per laboratory [2]. In our laboratory, the amount of teardrop cells is indicated with 1 + (which stands for 0.5–2% teardrop cells), 2 + (2–5%), and 3 + (>5%) [3]. To date, no cutoff value is established for the percentage of teardrop cells in a peripheral blood smear, discriminating pathology from nonpathology. If a reliable cutoff value can be determined for this red blood cell abnormality, this could lead to a faster and more accurate detection of myelofibrosis and in distinguishing this disease from benignity. How to realize a good, reliable, and automated screening tool for the detection of morphological red blood cell abnormalities (e.g., teardrop cells)? Automated morphological analysis and classification of leukocytes with the use of digital microscopy is nowadays a routine procedure in a large number of laboratories globally [4]. Recently, a novel software tool has been developed to detect and classify morphological abnormalities of red blood cells (advanced RBC application). The advanced RBC application (CellaVision) is basically a RBC cell-locating device. The system identifies and segments red blood cells. Each red blood cell is then characterized for size, shape, color, and inclusion. The characterizations for shape, color, and inclusion are performed by an artificial neural network. This network is trained by a number of highly qualified experts, and the network uses 80 features computed for each RBC image. Examples of used features are size, roundness, distribution of notches around the border, size, and shape of inner pallor. The user can verify or change any suggested morphology. The results are represented as a percentage value and a grading which is based on a conversion table defined by the user in the settings file. We evaluated and validated this module, using a cohort of patient samples and healthy controls, and compared preand postclassification results. Preclassification was performed by the RBC classification module without manual intervention (Figure 1), and the postclassification was performed by a morphological expert. Subsequently, statistical analysis was performed to determine the accuracy and correlation. Classification analysis of teardrop cells was performed on 46 peripheral blood smears from patients in which teardrop cells were present and a cohort of normal blood smears (n = 10). The slides were prepared using the SP10 (slide maker/stainer; Sysmex, Etten-Leur, the Netherlands) from venous blood samples collected in EDTA tubes and stained according to the May–Gr€ unwald– Giemsa stain. Of these 46 patients, fifteen were diagnosed with myelofibrosis, confirmed by bone marrow analysis. Myelofibrosis evolved from polycythemia vera (PV) in four patients and from essential thrombocythemia (ET) in five patients. Three patients were diagnosed with primary myelofibrosis (PMF) and three with idiopathic myelofibrosis (IMF). Among the other 31 samples, six were diagnosed with a myelodysplastic syndrome (MDS), four with chronic lymphocytic leukemia (CLL), three with an iron deficiency, one with ET, two with PV,

Interlaboratory Reproducibility of Blood Morphology Using the Digital Microscope

Differential counting of peripheral blood cells is an important diagnostic tool. However, manual morphological analysis using the microscope is time-consuming and requires highly trained personnel. The digital microscope is capable of performing an automated peripheral blood cell differential, which is as reliable as manual classification by experienced laboratory technicians. To date, information concerning the interlaboratory variation and quality of cell classification by independently operated digital microscopy systems is limited. We compared four independently operated digital microscope systems for their ability in classifying the five main peripheral blood cell classes and detection of blast cells in 200 randomly selected samples. Set against the averaged results, the R2 values for neutrophils ranged between 0.90 and 0.96, for lymphocytes between 0.83 and 0.94, for monocytes between 0.77 and 0.82, for eosinophils between 0.70 and 0.78, and for blast cells between 0.94 and 0.99. The R2 values for the basophils were between 0.28 and 0.34. This study shows that independently operated digital microscopy systems yield reproducible preclassification results when determining the percentages of neutrophils, eosinophils, lymphocytes, monocytes, and blast cells in a peripheral blood smear. Detection of basophils was hampered by the low incidence of this cell class in the samples.

Examination of peripheral blood smears: performance evaluation of a digital microscope system using a large‐scale leukocyte database

Sir, The analysis of blood morphology is of great diagnostic importance to the clinician. Manual morphological assessment using the microscope has been considered the gold standard for years but can be vulnerable to interobserver variability, is labor intensive, and requires highly and continuously trained personnel [1–3]. An exciting development in the field is the introduction of digital microscope (DM) systems. A DM ensures the constant presence of a morphological expert in the routine laboratory and enables the automated recognition of (pathological) cell types [3–5]. It was previously shown that the classification performance of the DM is equal to manual performance when classifying the five main peripheral blood cell classes (neutrophils, lymphocytes, monocytes, eosinophils, and basophils) [3–9]. However, these studies either used a low number of samples and cells or did not include a combination of normal and abnormal peripheral blood smears (PBS) [3–9]. Several studies also compared the postclassification results (which include manual interference) with manual analysis preventing a clear view on the ability of the DM to correctly classify cells without manual interference [4, 6, 7]. Here, we present a largescale database of about 1.4 million leukocytes from both normal and abnormal PBS, pitting the DM’s preclassification performance against the gold standard.

Digital microscopy as a screening tool for the diagnosis of hereditary hemolytic anemia

Evaluation of red blood cell (RBC) morphology is an important first step in the differential diagnosis of hereditary hemolytic anemia. It is, however, labor intensive, expensive, and prone to subjectivity. To improve and standardize the analysis of RBC morphology as a screening tool in the diagnosis of hereditary hemolytic anemia, we studied its automated analysis by digital microscopy (DM).

Rituximab addition to chemotherapy in real world patients with chronic lymphocytic leukemia: effective in first line but indication of lack of efficacy in subsequent lines of therapy

Lina van der Straten , Avinash G. Dinmohamed , Peter E. Westerweel, Anton W. Langerak, Jurgen Riedl, Jeanette K. Doorduijn, Arnon P. Kater and Mark-David Levin Department of Internal Medicine, Albert Schweitzer Hospital, Dordrecht, The Netherlands; Department of Research, Netherlands Comprehensive Cancer Organisation (IKNL), Utrecht, The Netherlands; Department of Public Health, Erasmus University Medical Centre, Rotterdam, The Netherlands; Department of Hematology, Erasmus MC Cancer Institute, Rotterdam, The Netherlands; Department of Immunology, Erasmus University Medical Centre, Rotterdam, The Netherlands; Department of Clinical Chemistry, Result laboratories, Albert Schweitzer Hospital, Dordrecht, The Netherland; Department of Hematology, Academic Medical Centre, Amsterdam, the Netherlands

The effectiveness of a routine versus an extensive laboratory analysis in the diagnosis of anaemia in general practice

Background We investigated the percentage of patients diagnosed with the correct underlying cause of anaemia by general practitioners when using an extensive versus a routine laboratory work-up. Methods An online survey was distributed among 836 general practitioners. The survey consisted of six cases, selected from an existing cohort of anaemia patients (n = 3325). In three cases, general practitioners were asked to select the laboratory tests for further diagnostic examination from a list of 14 parameters (i.e. routine work-up). In the other three cases, general practitioners were presented with all 14 laboratory test results available (i.e. extensive work-up). General practitioners were asked to determine the underlying cause of anaemia in all six cases based on the test results, and these answers were compared with the answers of an expert panel. Results A total of 139 general practitioners (partly) responded to the survey (17%). The general practitioners were able to determine the underlying cause of anaemia in 53% of cases based on the routine work-up, whereas 62% of cases could be diagnosed using an extensive work-up (P = 0.007). In addition, the probability of a correct diagnosis decreased with the patient’s age and was also affected by the underlying cause itself, with anaemia of chronic disease being hardest to diagnose (P = 0.003). Conclusion The use of an extensive laboratory work-up in patients with newly diagnosed anaemia is expected to increase the percentage of correct underlying causes established by general practitioners. Since the underlying cause can still not be established in 31.3% of anaemia patients, further research is necessary.

High coincidence of chronic lymphocytic leukemia and myeloproliferative neoplasms: detection bias or a clue to a common pathophysiological path?

With great interest we read the recent report ‘Chronic lymphocytic leukemia and myeloproliferative neoplasms concurrently diagnosed: clinical and biological characteristics’ by Todisco et al. describing a cohort of patients with coinciding diagnoses of chronic lymphocytic leukemia (CLL) and myeloproliferative neoplasms (MPN) in their center. The overall incidence of 13 patients with concomitant CLL in a cohort of 1719 PMN patients was noted to be higher than to be expected from the incidence of CLL in the general population. They found two patients with polycythemia vera (PV) and CLL. This lead the authors to speculate on a possible common pathophysiological basis for the occurrence of both a myeloid and lymphoid neoplastic clone in these patients.[1] Although we agree that the incidence of CLL observed in this MPN cohort is considerably higher than would be expected based on a calculated rate of coincidental occurrence, we wondered if a detection bias may confound the comparison with the general population. As CLL is present asymptomatically in the majority of patients, the prevalence is likely to be underestimated in the general population. This is apparent from other cohort studies of thoroughly medically screened subjects [2] and may explain the higher incidence of monoclonal B-cell lymphocytosis (MBL) in patients with MPN.[3] Asymptomatic CLL patients thus have a higher chance of detection of their CLL clone in case of a concomitant MPN diagnosis due to the investigations of peripheral blood and bone marrow they are generally subjected to. For comparison, we investigated the prevalence of concomitant myeloid malignancies in a retrospective cohort of 155 patients with PV from ten nonacademic teaching hospitals in the Netherlands we recently reviewed. We found one patient with a concomitant diagnosis of CLL, one patient with MBL and one patient with chronic myelomonocytic leukemia. The CLL patient in our cohort study was a 72-year-old female with JAK2 V617F positive PV in whom the diagnosis of CLL was made based on the routine immune fluorescence panel of both blood and bone marrow obtained in the context of investigations for her suspected PV at the time of diagnosis, prior to initiation of treatment. She had an asymptomatic Rai 0, Binet A stage CLL. Interestingly, there was a clear positive family history for PV in our patient as she had two relatives who were also diagnosed with PV. Secondly, we performed a chart review of all CLL patients diagnosed between 2000 and 2015 in our center. In the 374 CLL patients thereby identified, there were two patients with a concomitant MPN. One patient was the case from the PV cohort described above. However, there was a second case of a patient presenting with a combined diagnosis of CLL and MPL W515L mutation positive primary myelofibrosis. Our cases may be perceived as examples of detection bias. However, the consistency with which coincidental occurrences are observed in the various cohorts makes this less likely as CLL and MPN by themselves have a very low incidence. Also, the positive family history in our PV patient provides an important clue for an underlying genetic aberrancy predisposing to the development of mutations driving either myeloid or lymphoid clonal proliferation. Indeed, Todisco et al. also found a very high incidence (23%) of a positive family history in their MPN patients with CLL.[1] Taken together, we believe that although incidence rates of concomitant hematological