Image Classification of White Blood Cells with Canonical Correlation

Abstract

White blood cells are blood cells that are produced in the bone marrow and present in the blood and lymphatic tissue. They are a part of the immunological system of the body. They aid in the body's battle against infection and disease. WBC classification and identification can aid in the diagnosis of a variety of blood disorders.In the health-care industry, deep learning has numerous applications. One of the most promising areas of research is the classification of white blood cells. A tiny amount of blood is obtained and analyzed under the microscope to manually classify the WBC. This procedure takes a long time and has other drawbacks, including a lack of accuracy. Furthermore, there is the possibility that cells would overlap, which will impair accuracy. In this paper, we proposed a new method for identifying different types of White Blood Cells, such as monocytes, lymphocytes, eosinophils, and neutrophils. Convolutional neural networks are used to extract WBC features, and the CCA is used to remove nuclei from overlaying cells and learn from them. The proposed CNN with Canonical Correlation Analysis determines more accuracy than the previous approaches, according to experimental data.