Automated Blood Cell Classification Using Convolutional Neural Networks (CNN)

The accurate classification of blood cells from biomedical images is critical for diagnosinghematological disorders and advancing clinical research. This study presents a robust approachto automated blood cell classification using a Convolutional Neural Network (CNN) applied toa dataset of 5,000 images across five classes: basophil, erythroblast, monocyte, myeloblast, andsegmented neutrophil. The methodology integrates data augmentation techniques (random rotation, zoom, and flip) and a custom CNN architecture with convolutional, batch normalization,and dropout layers to enhance generalization and mitigate overfitting. The model was trainedon 4,000 images and validated on 1,000 images, achieving a validation accuracy of 96.0% anda loss of 0.127 after 21 epochs, with early stopping and learning rate reduction strategies optimizing performance. Evaluation via a normalized confusion matrix revealed high class-specificprecision, with diagonal values exceeding 0.89 for all categories. The trained model was savedfor future deployment, demonstrating potential for real-time blood cell analysis in medical diagnostics. This work highlights the efficacy of deep learning in processing biomedical images andits promise for improving diagnostic workflows.