A Novel and More Efficient Approach for Automatic Diagnosis of Acute Lymphoblastic Leukemic Cells based on Combining Geometrical and Statistical Features of Blood Cells

Acute lymphoblastic leukemia (ALL) is one of the most common types of leukemia among children. Due to the large number of clinical laboratories, in those with no expert pathologist for diagnosis of leukemia, software can be a useful tool for diagnostic purposes. The aim of this study was to create an automatic detector to help diagnosis process.
Using automatic segmentation algorithm, the nucleus of blast and lymphocyte cells were separated from existing images. As the chaotic characteristic caused significant difference in edges and string patterns, three geometrical, statistical, and chaotic features were derived from cells. In order to diagnosis and classification, support vector machine algorithm was used and the accuracy of classification was investigated using receiver characteristic operating curves (ROC).
Findings: This study was conducted on 312 microscopic images including blast and lymphocyte cells. There was a specificity of more than 92% and an accuracy of more than 93% in six cell groups. In addition, checking out the area under the ROC curve represented more than 91% efficiency for suggested method.
The findings indicate the effectiveness of these features in classification. Differentiation of blast and lymphocyte cells, that are different only in size of chromatin, and also uneven shape of lymphocyte cytoplasm, are of the advantages of using chaotic features.