Automated Detection Model in Classification B-Lymphoblast Cell from Normal B-Lymphoid Precursors in Blood Smear Microscopic Images Based on the Majority Voting Technique

Abstract

Introduction: Acute Lymphoblastic Leukemia (ALL) is a deadly white blood cell disease that affects the human bone marrow. Detection of ALL, the most common type of leukemia, has been always riddled with complexity and difficulty in its early stages. Peripheral blood examination as a common method at the beginning of the ALL diagnosis process is a time-consuming, tedious process and greatly depends on the experts' experience, keeping up with the advances in artificial intelligence in the diagnosis process. Keeping up with the growth and development of artificial intelligence algorithms a model was developed to classify B-ALL lymphoblast cells from lymphocytes. Materials and Methods: A Fast, efficient and comprehensive model based on Deep Learning (DL) was proposed by implementing eight well-known Convolutional Neural Network (CNN) models for feature extraction on all images and evaluating in classifying B-ALL lymphoblast and Normal. After evaluating their performance, four best-performing CNN models were selected to compose an ensemble classifier, by combining the model performance of each classifier. Results: Due to the close similarity of the nuclei of cancerous and normal blood B-ALL cells, the state-of-the-art CNN models alone did not achieve acceptable performance in diagnosing these two classes and their sensitivity was low. The proposed classification model Based on the majority voting technique was adopted to combine the CNN models. The sensitivity of 99.4, the specificity of 96.7, AUC of 98.3, and accuracy of 98.5 were obtained for the proposed model. Conclusion: To classify blood cancerous cells from normal cells, the proposed method can achieve high accuracy without the intervention of the operator in cell feature determination. Thus, the DL-based model can be recommended as an extraordinary tool for the analysis of blood samples in digital laboratory equipment to assist laboratory specialists.