China 
Africa 
Explore chapters and articles related to this topic
Systemic toxicology
Published in Chris Winder, Neill Stacey, Occupational Toxicology, 2004
W.M. Haschek, N.H. Stacey, C. Winder
Haematotoxicity can be targeted to cells of the bone marrow or those in the circulation (Irons 1985). Injury to the pluripotent stem cells or their microenvironment can result in underproduction of all cell types. Once differentiation begins, individual cell types can be affected, either in the bone marrow or in the blood, leading to a decrease in cell numbers of a particular cell type or abnormal function of that cell type. Direct toxicity to circulating cells results in an increased demand on the bone marrow resulting in increased cell proliferation (hyperplasia) and decreased differentiation time. Abnormal cell function will not be discussed in detail, since the target organs lie outside the haematopoietic system (see Bloom and Brandt (2002) for an in-depth discussion). A decrease in erythrocyte numbers, mean corpuscular volume (MCV), mean corpuscular haemoglobin content (MCH) or any combination thereof, is termed anaemia. A decrease in white blood cells is termed cytopenia (granulocytopenia, lymphocytopenia or thrombocytopenia depending on the cell type affected) or pancytopenia if all white cells are affected. Effects will vary in severity and reversibility, depending on the chemical and exposure conditions. Possible outcomes are recovery following cessation of exposure, persistence of changes, or progression to aplastic anaemia, myelodysplastic syndrome, or leukaemia (Rosner and Grunwald 1990).
Deep learning for few-shot white blood cell image classification and feature learning
Published in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2023
Despite the fact that AI models based on DCNN and transfer learning have obtained good accuracy for many applications of the blood cell classifications, many existing models carry numerous features extracted from deep neural networks, leading to high computational cost and memory requirements that prevent their deployment to clinic settings. In this work, we implement a deep learning model to automatically classify lymphocytes and non-lymphocytes with only a few image samples, which can be used to diagnose lymphocyte-induced WBC disorders, such as lymphocytic leukocytosis, a disease featured with abnormally high number of lymphocytes (Abramson and Melton 2000), and lymphocytopenia, a disease characterised by a reduced number of lymphocytes (Martin et al. 2017). In addition, we employed image denoising method to remove irrelevant but confusing image pixels around WBCs and observed the corresponding impact on the machine learning model performance. It is noted that under physiological conditions, neutrophils account for more than half counts of WBCs and lymphocytes taking up to 30% of the population of WBCs (Liu et al. 2020), which leads to the imbalance in the acquisition of WBC image samples. Under diseased states, the counting distribution of WBC subtypes may become more skewed and hence decrease the model performance. Therefore, we implemented three data augmentation methods to improve classification performance on the minority samples and analyse the corresponding effects.