Advanced biomedical imaging for identifying blood cell type: Integrating segmentation, feature extraction, and GraphSAGE model

The analysis of blood, including red blood cells (RBC) and different types of white blood cells (WBCs) plays a major role in the diagnosis of certain diseases. Automated segmentation of blood cells and their components can assist clinicians in effectively making diagnoses; however, it is quite challenging Objective: This study proposes a computerized approach to assessing the significance of biomedical imaging. It presents a framework for segmenting blood cells as well as their nuclei from the histopathological images of multiple datasets. Additionally, a custom algorithm is developed for blood cell counting. This study introduces two automated methods for WBC analysis, including image segmentation to distinguish between WBCs and RBCs, the nuclei of the WBC, and classifying WBC types using clinically important features. An effective segmentation approach with image preprocessing algorithms is developed for automatic counting of WBCs and RBCs. An improved GraphSAGE model is constructed to classify blood cells. Clinically relevant features are extracted from segmented WBCs and nuclei for a final dataset. Feature ranking analysis identifies optimal features and reduces dimensionality, aiding graph dataset construction based on data similarity. Our proposed model achieved an accuracy of 96.67 %. A comparative analysis with benchmark models is done to assess the effectiveness of the model. The explainability of the model is addressed to enhance the transparency of the diagnostic system and provide insight into the decision-making process. Leveraging the automated, simultaneous segmentation of blood cells and exploring their relationships for effective classification substantially helps to improve the reliability and applicability of this diagnostic system and aid clinicians.