White Blood Cell Classification Using SMOTE-SVM Method with Hybrid Feature Extraction and Image Segmentation Using Gaussian Mixture Model

White blood cells are crucial to the immune system. The irregular structure of white blood cells, along with the fact that each type has its unique structure, makes manual identification challenging. Manual identification is prone to errors due to medical personnel's subjectivity and fatigue from time and effort demands. A fast and accurate method for classifying white blood cells is needed, but challenges remain regarding the quality and quantity of samples for each cell type. This study proposes the use of SMOTE and SVMSMOTE to address the issue of data imbalance, as well as a combination of shape features (size, circularity, convexity, solidity) and convolutional autoencoder (CAE) for feature extraction, along with a Gaussian mixture model for nucleus segmentation. The study finds that, without using SMOTE or SVMSMOTE for data balancing, the proposed features are already sufficient to represent each cell type except eosinophils, achieving an accuracy of 92.4%, precision of 91.9%, recall of 92.3%, F1-Score of 92%, MCC of 0.862, and CEN of 0.1376 using a polynomial kernel. The worst results were obtained with the sigmoid kernel. The combined feature extraction (shape and CAE) outperformed individual methods. Shape alone achieved 86.8% accuracy, CAE alone 87.8%. Recall for eosinophil cells improved using SMOTE and SVMSMOTE.