A novel automated immune cell classification and cancer stage prediction system based on advanced machine learning techniques is proposed. To specifically detect and segment immune cells, the system uses YOLOv8, resulting in >98% accuracy in identifying neutrophils, lymphocytes, and monocytes. Using a Convolutional Neural Network (CNN), these segmented images are subsequently classified into cancer stages. The study uses robust preprocessing techniques, such as normalization and data augmentation, to achieve consistent performance against diverse clinical conditions. Reliability in immune cell detection and cancer stage classification is demonstrated by the system, with average precision, recall, and F1-score above 90%. Real-time analysis and visualization through a Streamlit-based web application allow clinicians and researchers to use actionable information. The approach addresses the limitations of traditional microscopy-based diagnostics and reduces human error while increasing diagnostic speed and early cancer detection. This capability of integrating machine learning models into clinical workflows represents a potentially scalable and accurate diagnostic solution in healthcare.