Enhancing Hematological Analysis: White Blood Cell Classification Using CNNs
Gunjan Shandilya, Vatsala Anand, Rahul Singh Chauhan, Hemant Singh Pokhariya, Sheifali Gupta · 2024
White blood cells, also known as WBCs, have a vital function in the immune system as they protect the body from infectious diseases and foreign intruders. Abnormalities observed in many types of white blood cells might serve as indicators of diseases such as leukemia, underscoring the importance of precise classification for diagnosis and monitoring. Prior studies frequently encounter reduced precision as a result of employing less important characteristics and concentrating on a smaller number of white blood cell (WBC) types, hence inflating performance measurements. This research tackles these problems by introducing an innovative method that integrates thorough pre-processing and data augmentation strategies to provide a comprehensive set of attributes. The model has been trained using a dataset of 12,500 augmented photos, which are grouped into four types: Eosinophil, Lymphocyte, Monocyte, and Neutrophil. The training process uses a Convolutional Neural Network (CNN) classifier. The approach obtained an impressive accuracy of $98 \%$, with an overall precision and recall of $97 \%$, showcasing its better accuracy and computational efficiency in comparison to existing state-of-theart methods. This emphasizes the capability of this method for accurate and effective categorization of blood cells in medical diagnosis.