Akut Lenfositik Löseminin Makine Öğrenimi Yöntemleriyle Otomatik Tespitine İlişkin Karşılaştırmalı Bir Çalışma

Abstract

Acute Lymphocytic Leukemia (ALL) is one of the most prevalent types of leukemia which has the risk of death of children is relatively higher than adults. The early diagnosis of this disease is crucial and it can be detected by examining the morphological changes of the blood cells. In this study, we exhibit a comparative study on the automatic classification and identification of the ALL with machine learning methodologies. Acute Lymphoblastic Challange Database (ALL-CDB) served by the Cancer Imaging Archive, which consists of 6500 digital microscopic pathology images from 118 subjects, is used. As the first step, the geometric features are extracted and after, the feature selection was performed with Principal Component Analysis (PCA). Finally, the classification process on the selected features was carried out by using Naive Bayes, k-Nearest Neighbor (k-NN), Linear Discriminant Analysis (LDA), Decision Tree (DT), Random Forest (RF), Support Vector Machine (SVM), and Multilayer Perceptron (MLP) neural network methods. The results between the methodologies have been analyzed in terms of accuracy, precision, recall, and F1-score metrics. According to the results, MLP gives the both highest accuracy and F1-score with 97% to classify the ALL cells for leukemia.