Cancer Cell Classification From Peripheral Blood Smear Data Using the YOLOv8 Architecture

Accurate and timely detection of cancerous cells from peripheral blood smears is critical for early diagnosis and effective treatment planning. In this study, we propose an automated system for the classification of blood cells using state-of-the-art deep learning techniques, primarily leveraging the YOLOv8 architecture. The system is trained on high-resolution peripheral blood smear datasets to accurately detect and classify four major blood cell types: eosinophil, lymphocyte, neutrophil, and platelet. In addition to YOLOv8, we conducted comparative experiments using YOLOv10 and YOLOv11 architectures to evaluate improvements in detection precision, inference speed, and robustness under varying imaging conditions. Our approach incorporates advanced data augmentation and preprocessing pipelines to enhance model generalization while maintaining low computational overhead. The proposed method achieves high classification accuracy, demonstrating its effectiveness in assisting medical professionals for faster and reliable diagnostics. The integration of YOLO-based detection models provides a real-time, scalable solution for automated hematological analysis, enabling early detection of abnormal cells and potentially improving patient outcomes. This framework can be further extended to other cytological image datasets, offering a generalized approach to automated cancer cell detection and classification in clinical practice.

Keywords: Peripheral blood smear, Cancer cell classification, YOLOv8, YOLOv10, YOLOv11, Eosinophil, Lymphocyte, Neutrophil, Platelet, Deep learning, Automated diagnostics.