Exploring dyserythropoiesis in patients with myelodysplastic syndrome by imaging flow cytometry and machine‐learning assisted morphometrics

Abstract

BACKGROUND: The hallmark of myelodysplastic syndrome (MDS) remains dysplasia in the bone marrow (BM). However, diagnosing MDS may be challenging and subject to inter-observer variability. Thus, there is an unmet need for novel objective, standardized and reproducible methods for evaluating dysplasia. Imaging flow cytometry (IFC) offers combined analyses of phenotypic and image-based morphometric parameters, for example, cell size and nuclearity. Hence, we hypothesized IFC to be a useful tool in MDS diagnostics. METHODS: Using a different-from-normal approach, we investigated dyserythropoiesis by quantifying morphometric features in a median of 5953 erythroblasts (range: 489-68,503) from 14 MDS patients, 11 healthy donors, 6 non-MDS controls with increased erythropoiesis, and 6 patients with cytopenia. RESULTS: First, we morphometrically confirmed normal erythroid maturation, as immunophenotypically defined erythroid precursors could be sequenced by significantly decreasing cell-, nuclear- and cytoplasm area. In MDS samples, we demonstrated cell size enlargement and increased fractions of macronormoblasts in late-stage erythroblasts (both p < .0001). Interestingly, cytopenic controls with high-risk mutational patterns displayed highly aberrant cell size morphometrics. Furthermore, assisted by machine learning algorithms, we reliably identified and enumerated true binucleated erythroblasts at a significantly higher frequency in two out of three erythroblast maturation stages in MDS patients compared to normal BM (both p = .0001). CONCLUSION: We demonstrate proof-of-concept results of the applicability of automated IFC-based techniques to study and quantify morphometric changes in dyserythropoietic BM cells. We propose that IFC holds great promise as a powerful and objective tool in the complex setting of MDS diagnostics with the potential for minimizing inter-observer variability.