Machine learning for thalassemia detection: a critical review of diagnostic pathways and emerging technologies

Thalassemia remains a major diagnostic challenge because its detection requires the interpretation of multiple laboratory and molecular findings in different stages of care. Current diagnostic approaches are effective in clinical practice, but they are often discussed as separate methods rather than as parts of a connected diagnostic pathway. This review addresses this need through a pathway-based and methodologically oriented perspective, beginning with traditional diagnostic methods, moving to emerging diagnostic technologies, and then examining machine learning (ML) as a supportive layer within the diagnostic pathway. To clarify the role of ML, the discussion separates methodological aspects from clinical applications: the first part explains how models are developed and evaluated, while the second part examines where they may contribute in real diagnostic practice. Finally, key challenges are discussed that must be addressed for reliable clinical translation and future implementation. Overall, this review argues that the future of thalassemia detection lies in an integrated diagnostic framework that connects screening, confirmation, monitoring, and treatment planning to support faster, more reliable, and more individualized care for a variety of populations.