White Blood Cells Classification: A Feature-based Transfer Learning Approach

Description

White blood cells (WBCs) are vital components of the immune system, playing a crucial role in defending the body against infections and diseases. Traditionally, the classification of WBCs was performed manually by hematologists, a method fraught with limitations such as subjectivity and time consumption. To address these challenges, this study explores the efficacy of a transfer learning pipeline for WBC classification, leveraging pre-trained models like VGG16, VGG19, and InceptionV3 for feature extraction. The classification models employed include k-Nearest Neighbour (kNN), Support Vector Machine (SVM), and Logistic Regression (LR). The research utilizes Paul Mooney’s white blood cell dataset, which consists of 640 microscopic images of four different WBC types. The dataset is divided into training, validation, and testing sets in a 70:15:15 ratio. The results demonstrate that the VGG16-LR pipeline achieves superior classification accuracy compared to other model combinations tested. This study highlights the potential of integrating transfer learning with robust classifiers to improve the accuracy and efficiency of WBC classification, offering promising implications for medical diagnostics and treatment planning.

Period	22 Aug 2024
Event title	International Conference on Intelligent Manufacturing and Robotics 2024
Event type	Conference
Location	Taicang, Suzhou, ChinaShow on map
Degree of Recognition	International