Color-based hybrid modeling to classify the acute lymphoblastic leukemia

Abstract

In this research, a new approach color based hybrid modeling to separate and classify the Acute Lymphoblastic Leukemia is proposed to resolve the classification tasks. A series of the process is offered to conduct the classification tasks, i.e., Pre-processing, Image Segmentation, Features Extraction, and Similarity Measurement. Furthermore, the Otsu thresholding-based contrast stretching model is proposed to enhance the image quality on the Pre-processing stage. Moreover, the largest object selection of the found objects indicates the right choice to capture the desired object. Furthermore, the feature extraction is performed by combining the color based densitometry and shape features. Lastly, the feature extraction results are measured by the Euclidian Distance and Manhattan method to classify the sample used, which is Acute Lymphoblastic Leukemia Image Database (ALL-IDB). The results of the proposed model have produced 95.38% accuracy. Therefore, it showed that the classification had produced higher accuracy than the others, i.e., Naive Bayesian, Color Correlation, Fuzzy-based Leukemia Detection, Hausdrof SVM-based Leukemia Detection, and Automated Differential (Learning Vector Quantization, Multi-Layer Perceptron, and Support Vector Machine.