Leukocyte Classification based on Transfer Learning of VGG16 Features by K-Nearest Neighbor Classifier
Diana Baby, Sujitha Juliet, M. M. Anishin Raj
Abstract
White blood cells (WBCs) are also called as leukocyte which is a significant component of blood that covers 1% of the total blood, protect us from numerous types of illness and other diseases. The automated classification of different types of leukocytes is very significant since each component have some designated functions in the human body and also the manual classification by skilled medical professionals is a tedious and erroneous task. In this work an automated approach based on transfer learning methodology is used for the detection and classification of leukocytes into four types such as Lymphocyte, Monocyte, Eosinophil and Neutrophil since there are limited numbers of images in the dataset. The methodology adopted in this work is a combination of deep learning and machine learning in which the features are extracted from the segmented nucleus of leukocyte by VGG16 deep learning model which is trained and evaluated using K-Nearest Neighbor (KNN) machine learning algorithm which provided an accuracy of 82.35% which is better when compared to Naive Bayes Classifier.