Supervised Acute Lymphocytic Leukemia Detection and Classification Based - Empirical Mode Decomposition
Reem Magdy Elrefaie, Elsaid A. Marzouk, Mohamed A. Mohamed, Mohamed Maher Ata · 2022 International Telecommunications Conference (ITC-Egypt) · 2022
Acute lymphocytic leukemia (ALL) is a potentially fatal disease that affects both children and adults. ALL spreads quickly in the human body and kills within weeks. Hematologists examine the blood and bone marrow to see if ALL is present. Manual blood testing methods, which have been used for a long time, might be slow. This study provided an advanced computerized mechanism for classifying leukemia into two types: normal cell leukemia and blast cell leukemia. To begin, advanced image preprocessing algorithms were used and applied to acquire data augmentation, enhancement, and transformation. Furthermore, the K-means clustering technique was used to accurately segment the appropriate nuclei from the background. Moreover, most salient features have been extract via an empirical mode decomposition (EMD) based Hilbert Hung Transform (HHT). The extracted features have been fed toward supervised classifiers in order to the classify the nuclei. A neural networks (NNs) classifier with the Bayesian regularization (BR) method have been implemented. Moreover, support vector machine (SVM), KNN, Random forest, Nave Bayes, Logistic regression, and Decision trees classifiers have been utilized, tested, and compared with NNs. Experimental results show that NNs outperformed a classification accuracy of 98.7%, sensitivity of 99.3%, and specificity of 98.1%.