Journal
EUROPEAN BIOPHYSICS JOURNAL WITH BIOPHYSICS LETTERS
Volume 50, Issue 8, Pages 1111-1127Publisher
SPRINGER
DOI: 10.1007/s00249-021-01574-w
Keywords
Cell morphology; Cell contour; Cell nucleus architecture; Fractal dimension; Lacunarity; Gray-level co-occurrence matrix (GLCM) features
Categories
Funding
- Institute of Physics Belgrade, University of Belgrade by the Ministry of Education, Science, and Technological Development (MESTD) of the Republic of Serbia
- MESTD of Serbia [451-03-68/2020-14/200015]
Ask authors/readers for more resources
Microscopic examination of stained peripheral blood smears is essential in evaluating patients with hematological and non-hematological diseases. Machine learning approaches are being used to improve prediction accuracy, and by analyzing a dataset of digitalized single-cell images, reliable parameters for WBC classification have been identified.
Microscopic examination of stained peripheral blood smears is, nowadays, an indispensable tool in the evaluation of patients with hematological and non-hematological diseases. While a rapid automated quantification of the regular blood cells is available, recognition and counting of immature white blood cells (WBC) still relies mostly on the microscopic examination of blood smears by an experienced observer. Recently, there are efforts to improve the prediction by various machine learning approaches. An open dataset collection including the recently digitalized single-cell images for 200 patients, from peripheral blood smears at 100 x magnification, was used. We studied different morphological, fractal, and textural descriptors for WBC classification, with an aim to indicate the most reliable parameters for the recognition of certain cell types. Structural properties of both the mature and non-mature leukocytes obtained from (i) acute myeloid leukemia patients, or (ii) non-malignant controls, were studied in depth, with a sample size of about 25 WBC per group. We quantified structural and textural differences and, based on the statistical ranges of parameters for different WBC types, selected eight features for classification: Cell area, Nucleus-to-cell ratio, Nucleus solidity, Fractal dimension, Correlation, Contrast, Homogeneity, and Energy. Classification Precision of up to 100% (80% on average) was achieved.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available