Publications

Bangladesh made substantial gains in child survival over two decades, with the under-five mortality rate falling from 94 to 30 per 1000 live births between 2000 and 2020. Recent data from the Bangladesh Bureau of Statistics, however, point to a troubling reversal. The infant mortality rate rose to 27 per 1000 live births in 2023, whereas neonatal mortality climbed from 16 to 20 per 1000 live births within a single year. More than 100,000 children under 5 died in 2023, with two-thirds of deaths occurring in the first month of life. Bangladesh also records around 63,000 stillbirths annually, representing a substantial public health burden and reflecting persistent gaps in maternal and antenatal care. Contributing factors include preterm birth, limited skilled attendance at delivery, pneumonia, malnutrition, and disruptions from the COVID-19 pandemic. Strengthening rural healthcare, expanding immunization, and improving maternal and neonatal care are essential if Bangladesh is to meet its Sustainable Development Goal (SDG) 3.2 targets by 2030.

Background: Dengue fever is a rapidly expanding vector-borne disease posing significant global epidemiological challenges. Accurate and interpretable forecasting is essential for timely interventions, yet most models overlook spatiotemporal, sex-specific, and country-level heterogeneity in disease dynamics.

Methods: A ConvLSTM network was applied to predict dengue incidence across 118 countries from 2000 to 2021 using 20 climatic, environmental, health system, and socio-economic predictors. Feature contributions were assessed through SHAP values, permutation importance, perturbation sensitivity, integrated gradients (IG), and layer-wise relevance propagation (LRP).

Results: ConvLSTM achieved the best predictive performance (R² = 0.7731). Annual freshwater withdrawals dominated male dengue incidence (SHAP: 44.37%), while hospital bed density had greater influence for females (SHAP: 31.86%). Temperature anomalies contributed consistently to both sexes (SHAP: 11.51%). Sensitivity analysis showed dengue incidence varied from −65% to +91% under ±50% predictor perturbations.

Conclusion: The ConvLSTM–XAI framework provides transparent, sex-aware, and country-specific dengue forecasts supporting targeted, climate-resilient dengue control strategies.

Background: AI, particularly ML and Deep Learning, is transforming healthcare by enabling improved diagnosis, prognosis, and personalized treatments. However, the opacity of many AI "black box" models limits interpretability, clinician trust, and real-world adoption. XAI has emerged to address these limitations by providing transparent and actionable insights. This systematic review synthesizes current evidence on XAI in healthcare, mapping AI models to XAI techniques, domains, and clinical applications.

Methods: A systematic search was conducted across six databases (Elsevier, Springer, Taylor & Francis, Semantic Scholar, ACM, and IEEE Xplore) for peer-reviewed articles published between 2017 and 2025, following PRISMA guidelines.

Results: Seventy studies were included, spanning oncology (40%), cardiology (21%), infectious diseases (14%), neurology (11%), and clinical decision support (13%). Deep learning models were most frequently applied (76%). SHAP (54%) and LIME (30%) were the most commonly used XAI techniques. Only 12 studies explicitly addressed ethical or regulatory considerations.

Conclusion: XAI enhances transparency, clinician trust, and decision-making in healthcare AI. Future work should focus on hybrid, clinically validated XAI models, comprehensive ethical compliance, and user-centered implementations.

Background: A life-threatening vector-borne disease, dengue fever poses significant global public health and economic threats, including Bangladesh. This study proposes an interpretable tree-based ML model for dengue early warning systems using climatic, sociodemographic, and landscape factors.

Methods: High-performance ML algorithms — Random Forests, XGBoost, and LightGBM — were applied to surveillance data from January 2000 to December 2021. SHAP values identified the most significant dengue drivers.

Results: Optimal thresholds: mean temperature 27°C, minimum temperature 22°C, humidity 82%. LightGBM accurately forecasts dengue, with agricultural land, population density, and minimum temperature as key outbreak drivers.

Conclusion: The proposed ML model functions as an early warning system, providing a framework for sophisticated analytical techniques in public health.