Beyond Diagnosis: A Systematic Review of Artificial Intelligence and Deep Learning in Monitoring Iron Overload and Organ Toxicity in Beta-Thalassemia

Transfusion-dependent β-thalassemia (B-TM) is complicated by progressive iron overload, remaining a primary cause of organ toxicity and mortality despite chelation therapy advances. Accurate monitoring of hepatic and cardiac iron levels is crucial. Traditional methods, like serum ferritin and magnetic resonance imaging (MRI) T2* relaxometry, improved outcomes but suffer from biological variability, operator dependence, and reduced precision in severe iron overload. Artificial intelligence (AI), machine learning (ML), and deep learning (DL) show promise in overcoming these limitations by automating image analysis, enhancing measurement precision, and enabling earlier detection of organ damage. This systematic review examines current evidence on AI and DL applications for monitoring iron overload and organ toxicity in B-TM patients. A search of PubMed, Scopus, and Web of Science identified studies (2019-January 2026) using AI-based models to evaluate liver iron concentration, myocardial iron concentration, or iron-related complications. Eight studies met the inclusion criteria. Most focused on DL-driven analysis of MRI R2*/T2* data for hepatic and cardiac iron assessment. These demonstrated improved segmentation accuracy, decreased interobserver variability, and better performance in severe iron overload. Additionally, ML models utilizing clinical data effectively predicted skeletal issues. While reported performance metrics were generally positive, most studies were retrospective, single-center, and lacked external validation, resulting in a high risk of bias. AI and DL approaches hold significant potential to transform iron overload monitoring in β-thalassemia. However, prospective multicenter validation, standardized reporting, and explainable AI are strictly required before these technologies can be adopted for routine clinical use.