Classify blood cell subtypes Using Deep Learning Techniques

Abstract

This work describes a technique for retrieving White blood cells (WBCs) employing blood microscopic images, identifying them, variety of WBC. A new challenge in the medical sector is diagnosing by comparing the number of WBCs with a normal number of them. A new problem in medical science is diagnosing by comparing the number of white blood cells (WBCs) to what is considered normal, and this topic has been studied. Image processing techniques are used in this study as promising modalities for diagnosing various types of blood malignant development in an effort to prevent these problems. Deciding the goal of the project is to employ image-processing methods to detect and classification of blood cell subtypes at an early stage. The proposed methodology is to detect and classify blood cell subtypes by implementing a TensorFlow library from the microscopic pictures of human blood cells using Convolutional Neural Network (CNN) architecture deep learning. After utilizing the TensorFlow framework and VGG16 architecture in our proposed methodology, we got 98.21% accuracy in training and 91.08% in validation individually. The TensorFlow model beat the Keras model with additional productive and precise outcomes.