Leveraging U-Net for Efficient Blood Cell Segmentation and Classification

Accurate segmentation and classification of blood cells are crucial for the diagnosis of various hematological disorders. Traditional methodologies often rely on manual analysis, which can be time-consuming and subject to human error. In this study, we introduce HemoDetect, an automated system that employs a U-Net architecture for the precise segmentation and classification of red blood cells (RBCs), white blood cells (WBCs), and platelets in microscopic blood images. The U-Net model was trained on a dataset comprising 360 annotated images, each paired with ground truth masks in XML format. The training process focused on enhancing segmentation accuracy, resulting in a validation accuracy of 92.83% and a validation loss of 0.1933 over 60 epochs. The dataset was divided into training and validation subsets using an 80–20 split ratio, ensuring robust evaluation of model performance. HemoDetect demonstrates significant advancements in automating blood cell analysis, with the potential to enhance diagnostic efficiency and accuracy in clinical settings. HemoDetect effectively segments and classifies blood cells, demonstrating a high level of accuracy that varies depending on the input image. The output visualizations comprise predicted segmentation masks overlaid on the original images, fa-cilitating clear interpretation of results. This system significantly enhances the efficiency of blood cell analysis, minimizing the need for manual examination and improving diagnostic capabilities. Future work will investigate the model's adaptability in detecting various hematological conditions, including anemia and leukemia, thereby broadening its clinical relevance.