Instance segmentation and contrastive clustering techniques for white blood cell analysis

White blood cell (WBC) analysis using artificial intelligence (AI) is growing rapidly. However, many available datasets lack labels, making supervised learning approaches challenging. This review uses both qualitative and quantitative methods. It includes peer-reviewed studies from Scopus indexed journals. Sources were selected from the Scopus. The review groups findings into three key areas: segmentation, clustering, and health scoring. Contrastive clustering enables grouping of cells without labels. New health scoring methods allow assessment of cell health and model confidence. Despite progress, challenges remain. Models often lack robust testing when ground truth data is unavailable. Ethical concerns and limited model explainability also pose barriers. Future research should aim to integrate segmentation, clustering, and clinical objectives within a unified framework.