Improved White Blood Cell Image Classification Using Deep Convolutional Neural Networks with Attention Mechanism

White blood cells (WBC) are a pivotal component of the immune system, with each subtype possessing unique functionalities in the body. Therefore, swift and accurate identification of WBC subtypes from microscopic images is of utmost importance in biological research. With the advent and progression of neural networks in machine learning, cell image classification through neural networks has gained widespread adoption. In this paper, a deep convolutional neural network (CNN) combined with attention mechanism is proposed to classify white blood cells in microscopic images. The pre-training of CNN on the ImageNet dataset helps to improve the training and convergence speed of the network, while enhancing the accuracy of the weight parameters. The attention mechanism enables the network to focus on important regions in the cell image, allowing for more accurate extraction of cell features and details. The model proposed in this paper is compared against other CNN models including ResNet, VGG16, Inception, etc. The results show that the model combining deep convolutional network and attention mechanism demonstrates superior stability in accuracy across various white blood cell image classification tasks. The introduction of the attention mechanism led to 88% accuracy in blood cell image classification, resulting in improved accuracy in the classification of white blood cells and effective utilization of critical information in the image.