Aug 2026· Frontiers in Digital Health· Vol 8· 0 citations· 44 references
Medicine
TL;DR
Artificial intelligence has considerable potential to improve NTD diagnosis in low-resource settings, but successful adoption depends on trust, transparency, and usability, so a proposed framework provides a structured pathway for developing explainable AI systems that are technically robust, clinically meaningful, ethically responsible, and implementable within resource-constrained health systems.
Abstract
Background Neglected tropical diseases (NTDs) continue to affect more than one billion people globally, disproportionately impacting populations living in low-resource settings characterized by limited diagnostic infrastructure, shortages of trained healthcare personnel, and restricted access to specialist services. Recent advances in artificial intelligence (AI), particularly deep learning and computer vision, have demonstrated significant potential for improving disease detection through the analysis of clinical images and microscopy data. However, despite encouraging diagnostic performance, many AI systems remain difficult to interpret, creating barriers to clinical trust, adoption, regulatory acceptance, and sustainable implementation in endemic regions. Main body This narrative review examines the current landscape of AI applications in NTD diagnosis and critically evaluates the role of explainable artificial intelligence (XAI) in addressing challenges associated with transparency and trustworthiness. Evidence from studies involving malaria, schistosomiasis, soil-transmitted helminth infections, leishmaniasis, and skin-related NTDs demonstrates the growing capacity of AI to support diagnostic decision-making in resource-constrained environments. Nevertheless, persistent challenges related to limited datasets, poor data quality, algorithmic bias, model drift, infrastructure constraints, and ethical governance continue to impede translation into routine healthcare practice. Existing explainability approaches, including Gradient-weighted Class Activation Mapping (Grad-CAM), heatmaps, Shapley Additive Explanations (SHAP), Local Interpretable Model-Agnostic Explanations (LIME), and attention mechanisms, were reviewed to assess their relevance for NTD diagnostic systems. Framework development Drawing upon current evidence in explainable AI, digital health implementation, and global health systems research, a seven-stage framework is proposed comprising: (1) problem definition, (2) data acquisition, (3) model development, (4) explainability layer integration, (5) clinical validation, (6) deployment in low-resource settings, and (7) continuous learning and monitoring. The framework embeds explainability throughout the AI development lifecycle to enhance transparency, accountability, clinical relevance, and equity. Conclusions Artificial intelligence has considerable potential to improve NTD diagnosis in low-resource settings, but successful adoption depends on trust, transparency, and usability. The proposed framework provides a structured pathway for developing explainable AI systems that are technically robust, clinically meaningful, ethically responsible, and implementable within resource-constrained health systems, thereby supporting future NTD control and elimination efforts.
Malaria endures a significant part in public health concern, especially in tropical and subtropical regions. Traditional malaria control methods often face limitations with surveillance, diagnosis and efficient resource allocation. This review explores the role of Artificial Intelligence (AI) in augmenting data-driven decision-making for malaria control and elimination efforts, focusing on surveillance systems, enhancing the effectiveness of intervention strategies and optimizing the resource allocations. AI technologies, mainly machine learning algorithms and computer vision systems, demonstrate significant potential in improving malaria control outcomes. Key findings include increased accuracy in outbreak prediction, improved diagnostic precision through automated microscopy and optimized resource allocation reducing response times. Additionally, deep learning models are emerging as promising tools in identifying drug resistance patterns and personalizing treatment protocols. AI integration in malaria control programs offers substantial benefits for public health decision-making. In this article, we conducted a comprehensive review of peer-reviewed literature, analyzing AI applications in malaria control across key domains such as surveillance, diagnosis, treatment and resource management. However, effective implementation requires robust data infrastructure, ethical frameworks addressing algorithmic bias and sustained international collaboration. Future directions prioritize equitable access, capacity building and development of standardized evaluation metrics for evaluating AI-driven interventions.
Sweta Bhan, Ayushi Singh, Pankaj U. Ramteke et al.· International Journal of Com...· 0 citations
Skin-manifesting neglected tropical diseases (NTDs) pose significant diagnostic challenges due to overlapping clinical presentations and limited access to specialist care in endemic regions. Artificial intelligence (AI) has shown promise in dermatological diagnosis; however, concerns remain regarding algorithmic bias, reduced accuracy in darker skin tones, and lack of transparency in decision-making. This review aimed to synthesise existing evidence on explainable AI approaches for the differential diagnosis of skin-manifesting NTDs, with emphasis on performance, equity across dark skin tones, and clinical applicability. A structured narrative review was conducted using systematic search methods across PubMed/MEDLINE, Scopus, AJOL, and ScienceDirect. Eligible studies included peer-reviewed AI-based diagnostic research involving skin conditions that incorporated explainability or interpretability methods. Literature published between 2015 and 2025 was screened and synthesised thematically. Evidence from studies demonstrated that deep learning models achieve high diagnostic performance in dermatology (often > 85% accuracy), but consistently underperform in darker skin tones, with reported reductions of up to 20%. Explainable AI techniques such as saliency maps, Grad-CAM, and confidence scoring were shown to enhance interpretability and support differential diagnosis, though limitations related to dataset diversity and real-world deployment persist. Explainable AI represents a critical advancement for equitable and reliable diagnosis of skin-manifesting NTDs. Addressing dataset bias, embedding transparency, and aligning AI tools with frontline workflows are essential to maximise clinical and public health impact. Not applicable.
D. C. Innocent, Precious Ebube Anyakorah, Rejoicing Chijindum Innocent et al.· BMC Artificial Intelligence· 0 citations
The application of artificial intelligence (AI) in plant disease detection has rapidly advanced, enabling accurate, timely, and automated diagnosis from image data. Deep learning models, particularly convolutional neural networks (CNNs) and hybrid architectures, have demonstrated high classification performance, often exceeding 95% accuracy under controlled conditions. However, the lack of model interpretability remains a major barrier to real-world adoption in agricultural settings. To address this challenge, explainable artificial intelligence (XAI) techniques have been increasingly integrated to enhance transparency and user trust. This scoping review systematically examines recent studies (2021–2025) on AI-based plant disease detection with a focus on model architectures, explainability methods, datasets, and practical applicability. A total of 17 studies were analyzed to identify key trends and research gaps. The findings indicate that Grad-CAM, LIME, and SHAP are the most commonly used XAI techniques, providing visual and feature-level explanations that support model interpretability. Despite these advancements, significant challenges persist, including high computational cost, lack of standardized evaluation frameworks, and inconsistencies in explanation quality. Furthermore, the reliance on limited and curated datasets restricts model generalization under real-world conditions. The review highlights a critical gap between high-performing experimental models and their practical deployment in agriculture. Future research should focus on developing diverse datasets, lightweight and interpretable models, and standardized evaluation strategies to enable robust, transparent, and scalable AI systems for sustainable agriculture.
Emmanuel Yahaya· International Journal of Mul...· 0 citations
Future directions encompass multimodal diagnostic integration, transfer learning with foundation models, large language model (LLM)-assisted decision-making decision-making, and a "human-animal-environment" intelligent prevention and control system.
Donghao Su· Theoretical and Natural Scie...· 0 citations
Background Sub-Saharan Africa continues to shoulder the heaviest burden of malaria. The 2024 WHO malaria report highlighted that Africa contributed an alarming 94% of the global cases and 95% of the deaths. In the WHO African region, progress towards elimination and management of malaria is hindered by weak health systems, and lack of traditional diagnostic methods such as microscopy and malaria rapid diagnostic tests (mRDT). The primary aim of the study is to develop a machine learning (ML) ensemble model for malaria diagnosis using clinical and demographic data, tailored for resource-limited settings. Methods A retrospective study was conducted using 637 patient records from Gutu Mission Hospital and Gweru Provincial Hospital in Zimbabwe. Clinical symptoms (fever, chills, abdominal pain, headache and diarrhea) and demographic features (age, gender, residence and travel history) were analysed. Data preprocessing included handling class imbalance using Synthetic Minority Oversampling Technique (SMOTE) and feature selection using Recursive feature elimination (RFE). Seven individual ML models including Logistic regression (LR), Random Forest (RF), Decision Trees (DT), Gradient Boosting (GB), K-Nearest Neighbor (KNN), Naive Bayes (NB) and XGBoost were trained and evaluated on the malaria dataset. The individual models were further combined to build, train and evaluate ensemble models such as Bagging, Stacking, Soft Voting and AdaBoost. Model performance was assessed using accuracy, precision, confusion matrices, recall and F1score and AUR-ROC metrics. Results Clinical symptoms (chills: p=0.001, fever: p=0.003, diarrhoea: p=0.01, abdominal pain: p<0.001) were statistically significant predictors of malaria. Of the demographic factors, only travel history (p=0.02) showed significant association with malaria. Among the seven individual ML models, GB achieved the highest predictive performance (Accuracy = 0.94), followed by RF (Accuracy = 0.94%) and XGBoost (Accuracy = 0.93%). The stacking ensemble model outperformed all individual ML models and other ensemble models (bagging, soft voting and adaBoost) achieving accuracy = 0.96, precision = 0.95, recall = 0.98, F1 score= 0.96 and AUC-ROC = 0.98. Conclusion This study demonstrates that ML particularly ensemble models can be used to significantly improve malaria diagnosis. The integration of these models into a web-based application could provide a scalable and accessible diagnostic tool for healthcare workers in resource limited settings.
Panashe Nyengera, H. Takawira, F. Mlambo· medRxiv· 1 citation
This framework illustrates, without any claim of clinical validity, how a leakage-safe ML pipeline and SHAP interpretability can be combined and rigorously self-audited; real patient-level data and external validation are required before any clinical inference is drawn.
David Chepkonga, A. Langat, Ebenezer Esenogho et al.· Asian Journal of Research in...· 0 citations