Children with beta-thalassemia major (β-TM) are at risk of neurodevelopmental or cognitive impairment. In this study, we developed SurfGNN, a surface-based graph neural network model, to estimate brain age from cortical morphological features extracted via structural MRI, to a cross-sectional cohort of 25 β-TM patients and 40 age- and sex-matched healthy controls (ages 6-15 years). A brain age index (BAI)-the difference between predicted and chronological age-was computed and analyzed in relation to cognitive performance (Wechsler Intelligence Scale scores) and hematological parameters. Results showed that β-TM children had significantly lower cognitive scores and more delayed brain age (lower BAI) compared to healthy controls. Furthermore, BAI significantly correlated with hemoglobin and ferritin levels and mediated their association with cognitive performance, as demonstrated by mediation analysis. The SurfGNN model outperformed conventional machine learning baselines and provided interpretable insights into regional cortical alterations. Our findings suggest that BAI may serve as a sensitive biomarker for detecting early brain developmental delays in β-TM.

Thalassemia is one of the most common and harmful single-gene recessive disorders in the clinic. It is characterized by impaired or absent production of one globin chain of hemoglobin in adults. The most common form is beta-thalassemia, which is associated with defects in the production of beta-globin chain, resulting in an imbalance in the ratio of alpha-globin to beta-globin. As a result, unbound free alpha-globin chains precipitated in red blood cell precursors, resulting in iron overload, ineffective red blood cell production, and changes in mesenchymal stem cells, osteoblasts, osteoclasts, and other associated cells. Among them, iron overload and ineffective red blood cell production will destroy the bone marrow microenvironment, which will cause further damage to the hematopoietic system. Therefore, the correction of the bone marrow microenvironment plays an important role in the treatment of thalassemia. This review summarizes studies on the destruction of the bone marrow microenvironment in beta-thalassemia, treatment options for improving the microenvironment and stem cell transplantation.

Hydrops fetalis is a fetal condition characterized by abnormal fluid accumulation in multiple fetal compartments and is associated with high perinatal mortality. Unlike high-income settings, both immune and non-immune hydrops fetalis (IHF and NIHF) remain relevant in low- and middle-income countries (LMICs), where evaluation and specialized care are often delayed or inaccessible. We report two cases of hydrops fetalis diagnosed antenatally at a tertiary referral hospital in Kenya, followed through delivery and outcome. Case one involved immune hydrops fetalis due to suspected non-Rhesus alloimmunization, presenting with fetal ascites and cardiac dysfunction. Management included paracentesis and intrauterine transfusions, which temporarily stabilized the fetus, but the neonate died on day two of life. Case two involved non-immune hydrops fetalis initially presenting with isolated pleural effusion, which progressed to generalized hydrops. Amniotic fluid testing showed Rubella IgM positivity. Management included thoracocentesis and conservative management till 34+0 weeks´ gestation. Unfortunately, the pregnancy resulted in an intrauterine fetal demise. These cases illustrate the wide spectrum of hydrops fetalis and the diagnostic and therapeutic challenges in resource-limited settings. While antenatal interventions may prolong gestation, outcomes are largely determined by the underlying cause and disease severity.

Thalassemia is a hereditary blood disorder characterized by abnormal hemoglobin production. Common diagnostic methods include complete blood count, high-performance liquid chromatography, and hemoglobin electrophoresis. While physicians make the final diagnosis, advancements in artificial intelligence, specifically machine learning (ML) and deep learning, offer significant potential as auxiliary tools and decision support systems to reduce diagnostic errors. This study investigates ML algorithms for classifying thalassemia and its subtypes, including alpha (α) thalassemia and beta (β) thalassemia (minor, intermedia, and major). A synthetic training dataset of 1534 samples was generated based on the statistical properties and correlation structures of real clinical data. The models were then evaluated using an external real-world dataset of 349 patients from the Hematology Department of Atatürk University Research Hospital. Support Vector Machines (SVM), Logistic Regression (LR), XGBoost, Artificial Neural Networks (ANN), and a hybrid stacking model named ThalP were implemented. The ThalP model integrates the probability outputs of SVM, LR, and XGBoost through a neural network meta-classifier. Experimental results demonstrate that the proposed ThalP model achieved strong performance on the real clinical dataset with an accuracy of 83.1% and a macro-F1 score of 0.80. These findings indicate that ML-based hybrid models can serve as effective decision-support tools for classifying thalassemia subtypes using routine hematological parameters.

In regions with a high prevalence of thalassemia, conventional diagnostic methods may fail to detect atypical or complex genetic variants. Whole-exome sequencing (WES) provides a comprehensive strategy to identify such variants, allowing more accurate genotype-phenotype correlation. Nevertheless, its optimal integration into clinical workflows and its incremental value over standard diagnostic approaches remain uncertain. This study aimed to evaluate the clinical utility of WES in a heterogeneous patient cohort and to propose a framework for its rational use in endemic settings. WES was performed on 21 patients with clinically suspected thalassemia from Huadu District, Guangzhou, stratified by severity. Variant analysis encompassed both coding and noncoding regions. On average, 5440.7 ± 94.3 insertions/deletions and 49,719.3 ± 492.5 single-nucleotide variants per individual were detected. Frameshift insertions/deletions were predominantly localized to the HBB gene. Noncoding single-nucleotide variants in 3' untranslated regions were associated with reduced mean corpuscular hemoglobin (HGB) concentration (301 ± 25 g/L in patients with vs 321 ± 18 g/L in those without 3' untranslated region variants, P = .008). Compound heterozygosity involving classical β-thalassemia mutations (CD41/42(-TTCT) β0 and IVS-II-654(C>T) β+) accounted for 85.7% of severe cases. Three rare non-β-globin variants were identified: MYB c.107G>A (p.R36H) in 1 case, a variant of uncertain significance potentially linked to fetal HGB regulation; HBD c.440A>T (p.H147L) in another, which produced an artificially normalized HbA2 value (3.1%), creating a risk of β-thalassemia misdiagnosis; and HBG1 c.364G>T (p.E122*) in a third case, co-occurring with α-thalassemia and iron deficiency. HGB levels declined significantly with increasing clinical severity (mild: 99.7 ± 2.5 g/L; very severe: 31.0 ± 26.1 g/L, P < .05). WES proved to be a powerful diagnostic tool, identifying causative or contributing genetic variants in all patients in this cohort, including those with complex or atypical presentations. Although it may not yet be feasible as a first-tier screening approach due to cost constraints, our findings strongly support the selective integration of WES into existing thalassemia diagnostic workflows in endemic regions. Such an approach enhances diagnostic precision, clarifies complex genotypes, and can guide more individualized management strategies.