PBC-NAS: neural architecture search for peripheral blood cells classification

Abstract

Peripheral blood cell (PBC) classification is crucial for identifying different types of blood cells and understanding their complex relationships that affect human health. PBCs include erythrocytes, leukocytes, and platelets, each with unique morphological and functional characteristics. Classifying these cells can help diagnose hematologic disorders and assess overall health status. Therefore, there is a growing need for automated blood cell methods to significantly improve the efficiency and accuracy of PBC classification. In this study, we propose a new neural architecture search method, namely PBC-NAS, to improve the accuracy and efficiency of PBC classification. The proposed method is compared with state-of-the-art methods and automatic neural architecture search methods, and it achieves better results in terms of classification performance and model complexity. PBC-NAS has achieved 2.4 points better average accuracy with 7.3 times fewer parameters than its closest competitor.