Eric Nyarko

Human Immunodeficiency Virus (HIV) is still one of the prominent global health concerns ever since its outbreak in the first half of the 1980s, accounting for more than 42.3 million AIDS-related de...

Chronic and acute wounds are a significant and increasing burden on health care systems worldwide. A key step in ensuring timely and evidence-based treatment planning and intervention is accurate, automated wound-type classification. Previous deep learning-based methods for wound classification have been mainly based on alone-employed convolutional image encoders or simple late-fusion methods combining image and location features. The approaches are not well explained and are not robust over a range of morphologically equivalent wound types. In this work, we introduce a multimodal, spatially-aware Graph Neural Network-Concept Bottleneck Model (GNN-CBM) framework. This architecture offers a combined reasoning about the appearance of the wound images and the context of the anatomical location, and provides an exposure of a bottleneck of latent clinical prototypes before final classification. We systematically compared the seven backbone architectures used in the visual encoding stage of the framework, such as CNNs (ResNet-50, EfficientNet-B0, EfficientNetV2-S, ConvNeXt-Tiny, DenseNet-121) and Vision Transformers (ViT-Base, Swin-Tiny) on the AZH Wound and Vascular Center dataset (Milwaukee, WI), which includes N = 311 unique patients and an augmented working set of 830 image-location pairs. In order to avoid data leakage, patient identifiers were used as grouping variable in a 5-fold stratified group cross-validation scheme, ensuring that patient images are not contained in more than one fold. Overall, Swin-Tiny outperforms the other backbones with the highest accuracy (0.809 ± 0.014), macro-F1 (0.805 ± 0.016) and mean AUC (0.954). Adjusted McNemar's testing (Bonferroni corrected across 6 comparisons) returned that Swin-Tiny performed significantly better than five of the six other backbones, except ConvNeXt-Tiny, which is not significant. An ablation study of the proposed Domain-Constrained Spatial Penalty Loss revealed that when the anatomical penalty term was removed, macro-F1 scores dropped from 0.805 to 0.792 for the best model, indicating the impact of the spatial-plausibility constraint. We discuss the mathematical formulation of our framework, the anatomical graph construction and its density statistics, the implications of unsupervised concept bottlenecks on real interpretability, hardware efficiency and highlight the importance of external, multi-center validation prior to clinical deployment.

Snakebite envenoming is a major neglected tropical disease disproportionately affecting rural populations in sub-Saharan Africa. In Ghana, evidence on the spatial and temporal distribution of risk remains limited, constraining targeted prevention and resource allocation. This study quantified district-level snakebite risk across Ghana, identified persistent hotspots and environmental drivers, and evaluated the relationship between snakebite burden and geographic access to treatment. Monthly district-level snakebite cases from Ghana's District Health Information Management System (2020 to 2025) were analyzed across all 261 districts using a Bayesian spatio-temporal model incorporating spatial effects, a temporal random effect, and a space-time interaction, fitted via Integrated Nested Laplace Approximation. Environmental covariates including rainfall, temperature, humidity, and NDVI quantified associations with risk. Relative risks, exceedance probabilities, Local Indicators of Spatial Association, and geographic accessibility identified priority districts. Snakebite risk showed strong spatial clustering and temporal variation. Persistent high risk districts were concentrated in Upper West (Daffiama Bussie Issa, Wa East, Wa West, Sissala East), Savannah (Bole, Gonja), North East (Mamprugu Moagduri), Western North (Bia East), Bono (Banda), Oti (Krachi Nchumuru), Western (Wassa East), and Eastern Region (Nsawam Adoagyiri, Fanteakwa North), though patterns evolved. Fanteakwa North emerged as the highest risk district nationally in 2025. Humidity and temperature were associated with increased risk, while rainfall and NDVI showed no significant effect. High risk districts often had poor treatment access, revealing inequities. This first nationwide Bayesian spatio temporal assessment provides an evidence base for surveillance, antivenom distribution, and interventions supporting WHO's 2030 snakebite reduction goals.

Dementia is a progressive neurological condition characterized by cognitive decline and structural brain changes that evolve. Longitudinal modeling of these changes is important for improving disease monitoring, identifying progression patterns, and supporting early risk stratification. This study developed an explainable longitudinal machine-learning framework for dementia progression, using cognitive and Magnetic Resonance Imaging (MRI)-derived biomarkers from the Open Access Series of Imaging Studies (OASIS-2) longitudinal dataset. The dataset included 150 subjects and 373 repeated observations classified as Non-demented, Demented, or Converted. Current-visit features, previous-visit features, and slope-based temporal features were constructed from Mini-Mental State Examination, Clinical Dementia Rating, normalized whole-brain volume, estimated total intracranial volume, atlas scaling factor, Age, and MRI delay. Baseline models were compared with a longitudinal gradient-boosted model, using patient-level splitting to reduce data leakage across repeated visits. The proposed longiGradient Gradient boosting model achieved the best held-out test performance, with an accuracy of 88.16%, a macro F1-score of 0.776, and a weighted F1-score of 0.860. The model showed strong classification performance for Demented and Non-demented individuals, while converted cases remained more difficult to identify. A regularized gradient boosting model was also evaluated as an overfitting sensitivity analysis; although it reduced the perfect training fit, it did not improve held-out test performance. Feature importance, permutation importance, and SHapley Additive exPlanations identified Clinical Dementia Rating as the dominant predictor, with slope-based Clinical Dementia Rating providing additional longitudinal information. These findings suggest that combining cognitive measures, MRI-derived biomarkers, and temporal feature engineering can improve dementia progression modeling, although external validation in larger longitudinal cohorts is needed.

Paediatric patients of snakebite envenoming are vulnerable to severe complications due smaller body size and dependence on adults for timely medical attention. However, there is limited literature on risk factors for mortality in pediatric populations, particularly within rural Nigerian settings, where snakebite envenoming is prevalent. A retrospective study was conducted using medical records at the Snakebite Treatment and Research Hospital (SBTRH) in Northeastern Nigeria, including paediatric patients aged 0 through 17 years who presented with a snakebite from January through December 2024. We estimated cumulative survival probability over time using Kaplan-Meier curves and identified predictors of mortality using multivariable Cox proportional hazards models. SBTRH managed 2,192 snakebite patients in 2024, out of which 723 paediatric patients were included in analyses; 702 of these patients were discharged alive, while 21 died in hospital. Those who took four hours or more to arrive at the hospital after being bitten were more likely to die in hospital compared to those who arrived in less than four hours (hazard ratio (HR) = 5.87, 95% CI = 1.15-29.95). Further, those who were not given antivenom (Echitab) were more likely to die in hospital than those who were given antivenom (HR = 10.07, 95% CI = 3.89-26.07). Aligning with other studies on poor outcomes following snakebite, this study identifies delayed hospital presentation and lack of antivenom administration among key predictors of in-hospital mortality among paediatric snakebite patients in Northeastern Nigeria. These findings reiterate a need for strengthening antivenom supply chains and implementing targeted educational campaigns.

Obesity is a major problem worldwide, particularly in Latin countries such as Mexico, Peru and Colombia. This study aims to evaluate five models (XGBoost, Random Forest, Multinomial Logistic Regression, Linear Discriminant Analysis (LDA) and Support Vector Machines (SVM)) from the statistical and machine learning fields to determine which of them is the most effective in classification of obesity levels in a multivariate classification model. The research uses a dataset of 2111 records consisting of 17 physiological and behavioral attributes which had to be preprocessed, such as one-hot encoding and multicollinearity filtering, and a powerful 10-fold cross validation methodology. It is shown that the results of ensemble methods are much better than those of the traditional parametric ones. XGBoost achieved the highest classification accuracy (97.54%), and an F_1-score of 0.9747, while Random Forest achieved the 2nd highest classification accuracy (95.64%). For all comparisons, McNemar's pairwise significance tests validated the superiority of XGBoost (with p<0.0001). Feature importance analysis was used to determine the most significant features as follows: weight, height, age and number of vegetables consumed per week (FCVC). While the study showcases the power of gradient boosting for capturing non-linear interactions in health data, the exclusive use of synthetic data (77% from SMOTE) poses drawbacks as it may overestimate performance and limit generalizability. The proposed ensemble architectures should be further evaluated on larger, all-nonsynthetic datasets to be clinically generalizable in the future.

Background: Snakebite envenoming (SBE) is a significant yet often overlooked public health crisis that primarily affects impoverished communities. Mobile applications (apps) can be effective tools for managing snakebites. To meet the World Health Organization’s (WHO) goal of reducing SBE by 50% by 2030, app developers must consider regional users’ preferences to ensure their apps provide relevant and accurate information. This study predicted the importance of various functionalities of a mobile app for managing snakebites based on user preferences. Our objective was to provide quantitative evidence on which functions should be prioritized to inform the development of a customized local app to enhance snakebite care in Ghana. Methods: A cross-sectional survey using a quantitative statistical experiment design method was conducted to identify healthcare workers’ preferences for vital mobile app functions for managing snakebites. Participants were selected from two deprived and predominantly rural districts in the Eastern region of Ghana through a multi-stage sampling technique. The attributes used in the questionnaire were developed based on literature reviews and focus group discussions, and a statistical block design was employed to create the choice tasks. To rigorously evaluate the performance of the machine learning (ML) models and reduce the risk of overfitting, we employed 5-fold cross-validation and multiple evaluation metrics. The data were analyzed using seven ML models. Results: The four most vital mobile app functions identified by the participants were “step-by-step assistance for victims and first responders” (utility estimates (β) =0.3924; 95% confidence interval (CI): 0.3183 to 0.4665), “providing educational and training materials” (β = 0.2243; 95% CI: 0.1500 to 0.2987), “identifying venomous snakes through clinical evidence or symptoms” (β = 0.1718; 95% CI: 0.0982 to 0.2454) and “identifying venomous snake biodiversity in your area/region” (β = 0.0898; 95% CI: 0.0176 to 0.1620). Conversely, the app functions that were less favored included “platform for sharing snakebite treatment experiences” (β = -0.2503; 95% CI: -0.3268 to -0.1738) and “designing educational games about snakes” (β = -0.3995; 95% CI: -0.4737 to -0.3252). These findings align with subgroup analyses by gender, suggesting a consistent understanding of needs across different demographic groups. Conclusions: This study provides quantitative evidence on which mobile app functions should be prioritized to inform the development of a customized local app to improve management and care for snakebite victims in Ghana and other sub-Saharan African countries.

Purpose: This study developed a Bayesian hierarchical spatio–temporal modeling framework to analyze factors and trends in malaria risk across Ghana′s 16 administrative regions from 2020 to 2024. The aim was to identify statistically significant areas with elevated or persistent malaria risk, to inform targeted intervention planning and support the National Malaria Elimination Program. Methods: This study utilized malaria incidence data from the Ghana Health Service′s District Health Information Management System–II covering the years 2020 to 2024. Meteorological data were sourced from the Visual Crossing Weather Data, and regional population estimates were obtained from the Ghana Statistical Service. To analyze the data, a Bayesian hierarchical spatiotemporal model with a Negative Binomial (NB) likelihood was implemented using Integrated Nested Laplace Approximation to account for overdispersion. The model included Conditional Autoregressive priors for structured spatial effects, first–order random walk priors for temporal dependence, and spatio-temporal interaction terms. Additionally, Local Indicators of Spatial Association (LISA) analysis with 999 conditional permutations was conducted to identify statistically significant spatial clusters, including high–high hotspots and low-low cold spots. Results: The NB model significantly outperformed the Poisson model, leading to a reduction in the dispersion statistic from 9,227.55 to 1.11. Humidity with a 1–month lag showed the strongest positive association with malaria risk, while the ultraviolet index had the greatest protective effect. Predictive relative risk maps identified persistent high–risk clusters in the northern and northwestern regions, specifically Upper West, Upper East, Bono, Ahafo, and Western North. LISA analysis indicated that Bono–Ahafo has been a stable high–high cluster from 2020 to 2023, while Ashanti has remained a consistent low-high anomaly. Additionally, Greater Accra and Central regions formed a significant low-low cluster in 2024. Conclusion: The Bayesian hierarchical spatio-temporal framework effectively characterized the complex transmission dynamics of malaria in Ghana. It revealed significant spatial dependence, temporal correlation, and interactions between these factors. By identifying persistent high–risk clusters and statistically significant spatial associations, this framework provides essential evidence to guide resource allocation. These findings support Ghana′s National Malaria Elimination Program Strategic Plan (2024–2028) by enabling targeted interventions in hotspots and optimizing the use of limited resources to sustain progress in low–transmission areas.