Leukocyte classification based on feature selection using extra trees classifier: a transfer learning approach

Abstract

The criticality of investigating the White Blood Cell (WBC) count cannot be underestimated, as white blood cells are an important component of the body's defence system. From helping to diagnose hidden infections to insinuating the presence of comorbidities like immunodeficiency, an accurate white blood cell count can contribute significantly to shaping a physician's assessment. The manual process performed by the pathologists for the classification of WBCs is a time consuming and tedious task, which is further disadvantaged by a lack of accuracy. This study concentrates on the automatic detection and classification of WBC without data augmentation into four subtypes such as eosinophil, monocyte, lymphocyte and neutrophil based on images from three different datasets. The methodology adopted in this paper is transfer learning approach in which the features are extracted using ResNet50, DenseNet121, MobileNetv2, Inceptionv3 and Xception deep learning models.The extra trees classifier is used as an intermediate stage for selecting most predominant features which reduce the execution time. When evaluating the performance on the basis of recall, precision, F-Measure and accuracy parameters, the classification of ResNet50 features selected by extra trees classifier using multi-class SVM provides the highest accuracy of 90.76% .

Thyroid Disease Prediction Using Selective Features and Machine Learning Techniques

“…Machine learning-based methods, especially ensemble techniques, are used to select the essential features. We have considered the extra tree classifier technique as one of the feature selection methods in this work [ 27 , 28 ]. The extra tree classifier randomly constructs multiple decision trees using the training dataset.…”

Developing Multiagent E-Learning System-Based Machine Learning and Feature Selection Techniques

“…In the test set, each one of the test nodes with each one of the trees is supported with a number of random features depending on each one of the decision trees. Each decision tree should select the relevant feature-based mathematical algorithm [ 26 , 27 ].…”

Modified U‐Net for cytological medical image segmentation

“…Using the YOLOv5 network, 22 propose a bone marrow cell detection by performing training to minimize of a new loss function based on penalizing erroneous predictions for similar classes. 23 exploit the idea of transfer learning; from several known networks, an extraction of the optimal features is performed through an extra trees classifier and a multiclass svm terminates the process to obtain a distribution of leukocytes in four classes. Transfer learning is again being used in the work of Bagido et al 24 where authors compare the pretrained models: Mobile NetV2, Vgg-16, Xception and InceptionResNetV2.…”

“…Another critical weakness is that it is difficult to distinguish between cells of the same lineage, that is, during their various stages of maturation. Using the YOLOv5 network, 22 propose a bone marrow cell detection by performing training to minimize of a new loss function based on penalizing erroneous predictions for similar classes 23 . exploit the idea of transfer learning; from several known networks, an extraction of the optimal features is performed through an extra trees classifier and a multiclass svm terminates the process to obtain a distribution of leukocytes in four classes.…”