Improving Blood Cell Subtype Classification Performance Using Data Augmentation with MobileNetV2
Anggi Muhammad Rifa’i, Mohd. Aboobaider Burhanuddin, Wahyu Hadikristanto, Alhadi Saputra, Dhani Ariatmanto, M. Shahkhir Mozamir · 2025
Blood cell subtype classification plays a crucial role in medical diagnostics, particularly in identifying hematological disorders. This study explores the application of data augmentation and transfer learning techniques using the MobileNetV2 architecture to enhance classification accuracy. A dataset comprising four blood cell subtypes Eosinophil, Lymphocyte, Monocyte, and Neutrophil was utilized. Data augmentation techniques, including preprocessing and image transformations, were employed to mitigate overfitting and improve model generalization. The MobileNetV2 model, pretrained on ImageNet, was fine-tuned to classify the subtypes effectively. The proposed model achieved a accuracy of 9 8,68% with a corresponding loss of 4,72%, showcasing the model's robustness. The findings highlight the potential of leveraging transfer learning and data augmentation to advance the automation of blood cell analysis, paving the way for more efficient and accurate diagnostic tools in clinical settings.