Computerized texture analysis of atypical immature myeloid precursors in patients with myelodysplastic syndromes: an entity between blasts and promyelocytes

Abstract

BackgroundBone marrow (BM) blast count is an essential parameter for classification and prognosis of myelodysplastic syndromes (MDS). However, a high degree of cell atypias in bone marrow hemopoietic cells may be found in this group of clonal disorders, making it difficult to quantify precisely myeloblasts, and to distinguish them from promyelocytes and atypical immature myeloid precursors. Our aim was to investigate whether computerized image analysis of routine cytology would help to characterize these cells.MethodsIn May-Grünwald-Giemsa stained BM smears of 30 newly diagnosed MDS patients and 19 cases of normal BM, nuclei of blasts and promyelocytes were digitalized and interactively segmented. The morphological classification of the cells was done by consensus of two observers. Immature granulocytic precursors, which could not be clearly classified either as blasts or promyelocytes, were called "atypic myeloid precursors". Nuclear morphometry and texture features derived from the co-occurrence matrix and fractal dimension (FD) were calculated.ResultsIn normal BM, when compared to myeloblasts, nuclei of promyelocytes showed significant increase in perimeter and local texture homogeneity and a decrease in form factor, chromatin gray levels, Haralick's entropy, inertia, energy, contrast, diagonal moment, cluster prominence, the fractal dimension according to Minkowski and its goodness-of-fit. Compared to normal myeloblast nuclei, the chromatin texture of MDS myeloblasts revealed higher local homogeneity and goodness-of-fit of the FD, but lower values of entropy, contrast, diagonal moment, and fractal dimension. The same differences were found between nuclei of normal promyelocytes and those of MDS. Nuclei of atypical myeloid precursors showed intermediate characteristics between those of blasts and promyelocytes according to the quantitative features (perimeter, form factor, gray level and its standard deviation), but were similar to promyelocytes according to the texture variables inertia, energy, contrast, diagonal moment, cluster prominence, and Minkowski's fractal dimension.ConclusionBM atypical immature myeloid precursors are difficult to be correctly classified in routine cytology. Although their cytoplasm is more similar to that of myeloblasts, computerized texture analysis indicates a nuclear chromatin remodeling more close to the promyelocyte, thus indicating an asynchronous intermediate maturation stage between blast and promyelocyte.

Fractal dimension and Shannon's entropy analyses of the architectural complexity caused by the inflammatory reactions induced by highly crystalline poly(vinyl alcohol) microspheres implanted in subcutaneous tissues of the Wistar rats

“…Fractal Dimension (FD), as defined by the fractal geometry theory, is an alternative and attractive tool to solving impossibilities presented by classical Euclidian geometry in describing irregular and discontinuous shapes observed in the nature, specially problems involving structural complexities and irregularities in biological systems 7, 8. Representative examples of FD analysis application can be found in different areas such as ophthalmology,9 molecular biology, neuropathology, physiological modeling,10, 11 dentistry,12 neoplastic tissues analyses,13–16 bone regeneration evaluation,17, 18 cellular rejection in postcardiac transplantations,19 and structural complexity analyses of blasts and promyelocytes present in myelodysplastic syndrome 20…”

Automatic Bone Marrow Cellularity Estimation in H&E Stained Whole Slide Images

“…[Color figure can be viewed at wileyonlinelibrary.com] matrices features, which has shown promising results (15)(16)(17)(18). For this purpose, previous studies have proposed using intensity and texture features for tissue classification within WSI H&E stains, for example, based on gray level cooccurrence matrix (GLCM), Gaussian Markov random field, and run-length The columns show original subimages, ground truth segmentations, results after the tissue classification, and results after the bone marrow segmentation, respectively.…”

“…(i) an example containing fibrin (tissue to the left), fibrosis (middle tissue), and bone (tissue to the right). [Color figure can be viewed at wileyonlinelibrary.com] matrices features, which has shown promising results (15)(16)(17)(18). Therefore, such features may be useful to classify between different tissue types in bone marrow biopsies.…”

Quantitative distinction of the morphological characteristic of erythrocyte precursor cells with texture analysis using gray level co‐occurrence matrix

“…However, which have been yet defined by qualitative descriptions, 3 which is also big barrier to teach the definition to the artificial intelligence to develop an automatic examination system. Recently, automatic segmentation and distinction systems are reported for lymphoid 4 or myeloid 5 cells with the geographical, color, and the texture futures of these cells.…”