Adrian Michalski, Bogumiła Kupcewicz

Automatic classification of peripheral blood smear cells by the example of lead poisoning
2020-07-28

Automatic classification of peripheral blood smear cells by the example of lead poisoning

The first symptoms caused by heavy metals poisoning are usually non-specific, therefore their diagnosis requires specialized knowledge and experience. Incorrect diagnosis can lead to various disorders and irreversible changes in patient’s health. Lead poisoning is one of the heavy metals poisoning which is associated with nonspecific symptoms and may cause a broad range of biochemical, physiological and behavioral disfunctions. Lead is commonly found in industry and manufacturing and in 2017, lead poisoning caused over a million deaths worldwide. Due to nonspecific symptoms, lead poisoning is often diagnosed too late or incorrectly. Early symptoms are headache and stomachache, which potentiate with metal concentration. Lead poisoning increasing risk of anemia due to inhibition the ability to produce hemoglobin by interfering the heme biosynthesis pathway and decreasing red blood cell survival. Because of regularly occurred disorders of hematopoiesis and the presence of atypical cell in smears, diagnosis of lead poisoning that gives relevant information is bone marrow and peripheral blood smear study. The manual smear test made by qualified technician involves the evaluation of the preparation by using an optical microscope. Manual microscopic examination of blood cells is time-consuming and subjective. Automation of this process would enable to reduce the risk of human failure and solve the problem with lack of professional staff. Traditional machine learning is the most popular approach to automating microscopic smear testing. This method consists of the following stages: microscopic image acquisition, preprocessing, cell segmentation, feature extraction, followed by classification into types, artifacts and atypical cells. This work presents the recent methods proposed to automate the analysis of peripheral blood and bone marrow smears using traditional machine learning methods. A review of different machine learning methods was carried out, focusing on the presentation of an algorithm for the construction of automatic blood cell classifiers.

Keywords: image processing, machine learning, peripheral blood smear analysis, leukocyte classification.

© Farm Pol, 2020, 76(6): 318–323