Elsevier

Leukemia Research

Volume 69, June 2018, Pages 54-59
Leukemia Research

Research paper
Inter-observer variance and the need for standardization in the morphological classification of myelodysplastic syndrome

https://doi.org/10.1016/j.leukres.2018.04.003Get rights and content

Highlights

  • Nearly 100,000 hematopoietic cell morphology from 499 MDS patients were evaluated.

  • Inter-observer variance in the assessment of dysplastic features of MDS was found.

  • Observers didn’t recognize hypo–granularity unless almost all granules are absent.

  • Pseudo Pelger–Huët anomaly was overlooked except for morphologically typical cells.

Abstract

In this era of genome medicine, the sub-classification of myeloid neoplasms, including myelodysplastic syndrome (MDS), is now supported by genetic testing in selected cases. However, as the initial suspicion and primary diagnosis of the disease still largely relies on morphological features and numbers of hematopoietic cells, the establishment of a uniform diagnostic basis, especially for cell morphology, is essential. In this study, we collected nearly 100,000 hematopoietic cell images from 499 peripheral blood smear specimens from patients with MDS and used these to evaluate the standardization of morphological classification by medical technologists. The observers in this study ranged between two to eleven for each image, and the images were classified according to MDS criteria through a web-based system. We found considerable inter-observer variance in the assessment of dysplastic features. Observers did not recognize cytoplasmic hypo–granularity unless almost all granules in neutrophils were absent. Pseudo Pelger–Huët anomalies were also often overlooked, except for cells with a very typical “pince-nez” appearance. Taken together, this study suggests a requirement for further standardization in terms of morphological cell classification, and a need for the development of automatic cell classification–supporting devices for the accurate diagnosis of MDS.

Introduction

Myelodysplastic syndrome (MDS) is a hematopoietic malignancy characterized by morphological atypia (dysplasia) of hematopoietic cells, ineffective hematopoiesis, peripheral cytopenia and progression to acute myeloid leukemia (AML). The disease is often found in the elderly: The median age of disease diagnosis is 76 years, with around 30 out of a 100,000 people aged 70 years or older diagnosed with MDS each year, whereas the frequency is 3–4 among people less than 70 years of age [1]. Recently, it was found that the disease becomes obvious when hematopoietic stem cells sequentially acquire several to ten somatic and/or germline gene mutations, and that gene mutations are frequently associated with disease phenotype or prognosis [2]. Based on these recent advances in the understanding of the molecular pathogenesis of the disease, the 2016 version of the WHO classification for myeloid neoplasms widely adopted gene mutation patterns for disease classification [3].

However, needless to say, the initial diagnosis of MDS is still, in principle, based on cell morphology and numbers, in which the disease is diagnosed by the following criteria: the ratio of myeloblasts is less than 5% in the bone marrow and less than 1% in the peripheral blood; and cytopenia or dysplasia is seen in at least one lineage [3], although the patterns and degree of dysplasia in peripheral blood cells, per se, seem not to be associated with an MDS subtype [4]. Nevertheless, morphological discrimination between normal and MDS cells is often very difficult, and thus requires substantial experience. For example, though dysplasia of MDS is defined as having a morphological abnormality in 10% or more of hematopoietic cells in a smear sample [3], other diseases or reactive conditions, such as various infections by pathogens or the administration of drugs, including anticancer agents, sometimes display dysplastic features resembling MDS [5], [6]. Hematological disorders, including aplastic anemia and primary myelofibrosis, also occasionally give rise to cells morphologically close to MDS. Furthermore, a certain number of healthy people exhibit dysplasia in hematopoietic cells [7]. Therefore, subtle changes in cell morphology may dissuade clinicians and medical technologists from a confident judgement [8], possibly giving rise to the variance observed in MDS diagnoses among observers and hospitals [9], [10].

In this setting, the board for the standardization of cell morphology in the Japanese Society for Laboratory Hematology (JSLH) is working to standardize the definition of dysplasia. In Japan, Matsuda et al. proposed classifying dysplasia into two categories: (groups A and B), in which group A includes dysplastic features highly characteristic of MDS, including a pseudo Pelger–Huët anomaly, cytoplasmic hypo–granularity (a decrease of cytoplasmic granules by more than 80% compared to normal granular cells), micromegakaryocytes and ring sideroblasts; while group B includes other dysplastic features less specific to MDS [11]. This categorization was also adopted by the study group for idiopathic hematopoietic disorders in the Japanese Ministry of Health, Labour and Welfare (MHLW). Likewise, the International Working Group on the Morphology of MDS (IWGM–MDS) proposes that the presence of a pseudo Pelger–Huët anomaly, cytoplasmic hypo–granularity (decrease of granules by more than two-thirds), abnormal chromatin clumping and macrocytes in granulocytes should be taken into account for dysplasia in MDS [12].

Of course, the quality of smear samples impacts on the judgement of dysplasia. Factors that determine the quality of samples include the application of peripheral blood or bone marrow specimens to glass slides, including their drying and staining, and the camera settings for taking cell images (when stored as digital files and observed through a display monitor). As for sample staining, according to our experience, the Wright–Giemsa staining method tends to stain cytoplasmic granules of neutrophils weakly compared to May–Grünward–Giemsa staining. Thus the former staining may lead to overlooking a decrease in the number of cytoplasmic granules. Collectively, it is very important to standardize the procedure for the preparation of smear samples among institutes/hospitals. The systematic education of clinicians and medical technologists is also crucial so that they are able to evaluate the dysplasia of MDS appropriately. In Japan, medical technologists for hematology are certified by the JSLH, with 1234 medical technologists have been certified as of 2017. However, certified technologists are scarce, especially in regional cities; thus, clinicians and medical technologists who lack relevant experience often have no choice but to evaluate the dysplasia of MDS without input from more experienced personnel.

Under these circumstances, we have developed a cell classification system that can assist in the identification of MDS cells present in peripheral blood. This project is ultimately aimed at the implementation of software that distinguishes subtle dysplasia present in MDS cells utilizing a machine learning concept. Although the automatic identification of dysplastic cells from blood specimens using an automated hematopoietic cell counting system has been attempted previously [13], [14], as far as we know, a system that identifies such cells using machine learning is lacking. To this end, we first obtained a total 97,924 cell images from 499 peripheral blood smear samples from patients with MDS (note, however, that the images unavoidably included non-malignant myeloid and non-myeloid cells because they were automatically taken by an imaging device, as described in Section Materials and methods). Each image was subsequently evaluated by at least two medical technologists (2–11, mean 5.67) who majored in hematology, and the data used as a “guide” for machine learning. However, we had previously found variation among observers for about 40% of the images. In this manuscript, we first disclose the evaluations entered by observers and discuss the issues that need to be overcome to realize an automatic diagnostic decision support system.

Section snippets

Sample preparation and obtaining digital images of peripheral blood cells

A total of 499 peripheral blood smear samples from patients with MDS, which had been prepared for clinical usage at each hospital that participated in this study, were anonymized and collected at the Department of Laboratory Medicine in Kumamoto University Hospital. MDS diagnoses were made at each hospital. In this study, we put our focus on the evaluation of dysplasia in individual hematopoietic cells. Therefore, we collected samples regardless of the MDS subtype, so that we could collect as

Overall classification of cell images

Among the 97,924 cell images independently classified by at least two observers, 880 images were classified as “platelet”, “no cell” or “non-typable” by more than 80% of observers. Although we omitted such images before providing them to observers, inappropriate images still remained because we initially manually assembled the series of cell images. After excluding images not suitable for this study, we finally chose 97,044 images for consideration.

We first explored the distribution of

Discussion

MDS is a hematological malignancy that develops based mainly on the age-related accumulation of somatic gene mutations in hematopoietic stem cells. In recent years, genetic pathways that determine MDS development have been revealed owing to the rapid development of high-throughput parallel sequencing technology. A hematopoietic stem cell that initially acquired a mutation in the genes encoding epigenome regulators gives rise to a status of clonal hematopoiesis of an indeterminate potential,

Conflict of interest disclosure

The authors of this paper declare they do not have any competing financial interests to disclose.

Acknowledgements

We would like to thank Ms. S. Sakata for excellent technical assistance as well as for clerical work. This study was partly supported by the Challenge Programme for ICT-Innovation provided by the Japanese Ministry of Internal Affairs and Communications.

References (24)

  • N. Gangat et al.

    Myelodysplastic syndromes: contemporary review and how we treat

    Am. J. Hematol.

    (2016)
  • B.B. Ganguly et al.

    Impact of chromosome alterations, genetic mutations and clonal hematopoiesis of indeterminate potential (CHIP) on the classification and risk stratification of MDS

    Blood Cells Mol. Dis.

    (2017)
  • Cited by (26)

    • Whole slide image representation in bone marrow cytology

      2023, Computers in Biology and Medicine
    • A geno-clinical decision model for the diagnosis of myelodysplastic syndromes

      2021, Blood Advances
      Citation Excerpt :

      The diagnosis can be challenging, especially in patients with pancytopenia with minimal dysplasia, and is subject to interobserver variability, with up to 30% to 40% disagreement in diagnosis.2 This difficulty is magnified for patients in whom the expected karyotypic or morphologic bone marrow changes are absent, or in the case of unsuccessful biopsies (eg, due to marked hypocellularity or extensive myelofibrosis).3,4 In addition, there is increasing recognition of the relationship between MDS, myeloid neoplasms, and the states of idiopathic cytopenia of undetermined significance (ICUS) and clonal cytopenia of unknown significance (CCUS).5

    • A new convolutional neural network predictive model for the automatic recognition of hypogranulated neutrophils in myelodysplastic syndromes

      2021, Computers in Biology and Medicine
      Citation Excerpt :

      This difficulty causes a considerable inter-observer variance in their manual classification. Therefore, the identification of cytoplasmic hypo-granularity becomes inconsistent among pathologists [6]. There is a clear need for further standardization of morphological cell classification and the development of computational classification models for the automatic detection of cytoplasmic hypogranulation for MDS diagnosis.

    View all citing articles on Scopus

    A commentary will follow in the upcoming issue for this article.

    View full text