Blood Platelets Detection Using YOLO v9 And XAI Tool

Blood platelets are measured or identified when evaluating haematological disorders like thrombocytopenia, thrombocytosis, and platelet function disorders. Though the methods of categorized counting are accurate, the time consumption and associated error rates make implementing automated methods desirable. This work proposes a novel approach using the YOLO v9 deep learning model for blood platelet detection and categorization on thin film blood smear images. Since YOLO v9 is one of the fastest and most precise versions of YOLO, it can detect fine objects like platelets in a demanding background. To support usability and reliability, this work includes Explainable Artificial Intelligence (XAI) tools, which feature regions of interest, feature importance, and prediction confidence. It empowers medical professionals to use artificial intelligence without fear, aiding ethical healthcare practices. This system introduces high precision and recall to reduce false results regarding the presence or absence of platelets. XAI can help bridge the gap between artificial intelligence and domain knowledge by allowing clinicians to verify the model’s outputs. Testing the system with different datasets shows significant potential for advancing platelet analysis.This work proposes a fervent, meaningful, and open approach to haematological diagnosis through an automatic analytical system. It is demonstrated that tuned detection models combined with interpretability frameworks enable significant progress in deploying AI in medical imaging and diagnostic workflows.