Evaluating the diagnostic accuracy of artificial intelligence-based systems for malaria detection in Africa: a systematic review and meta-analysis protocol

Malaria remains a major public health challenge in Africa, with 249 million cases and 608,000 deaths reported globally in 2022, predominantly affecting sub-Saharan Africa. While microscopy is the diagnostic gold standard, it is labour-intensive and expertise-dependent. Artificial intelligence (AI)-based systems, particularly convolutional neural networks, offer promising alternatives for automated malaria detection with potential advantages in accuracy, efficiency, and scalability. This systematic review protocol outlines methods to evaluate the diagnostic accuracy of AI-based systems compared to conventional microscopy for malaria detection in African populations. We will conduct a comprehensive search of PubMed, Web of Science, Scopus, and Cochrane Library for studies published between 2000 and 2025, supplemented by grey literature and reference list searches. Eligible studies will include observational studies, diagnostic accuracy studies, and randomized controlled trials that evaluate AI-based malaria diagnostic systems (including deep learning, machine learning, and computer vision approaches) against conventional microscopy or rapid diagnostic tests in African populations. Two independent reviewers will screen titles, abstracts, and full texts using predefined eligibility criteria, with a third reviewer resolving disagreements. Data extraction will capture study characteristics, population demographics, intervention details, comparator specifications, and diagnostic performance metrics including sensitivity, specificity, positive and negative predictive values, likelihood ratios, and turnaround time. Risk of bias will be assessed using Cochrane RoB 2.0 for randomized trials, QUADAS-3 for diagnostic accuracy studies, and Newcastle-Ottawa Scale for observational studies. Where appropriate, random-effects meta-analysis will pool studies evaluating different AI models (CNN architectures, machine learning classifiers, and computer vision systems) using bivariate or hierarchical summary ROC (HSROC) models, which accommodate between-study variability arising from differences in model architecture, training datasets, and implementation setting s. Subgroup analyses will explore heterogeneity by AI system type, microscopy technique, population characteristics, African region, and study quality. Sensitivity analyses will assess robustness by excluding high-risk studies and small sample sizes. The certainty of evidence will be evaluated using the GRADE approach for diagnostic test accuracy. This systematic review will provide comprehensive evidence on the diagnostic performance of AI-based systems for malaria detection in Africa, informing clinical practice, policy decisions, and future research priorities for implementing innovative diagnostic technologies in resource-limited settings. Where appropriate, random-effects meta-analysis will be used to calculate pooled estimates of diagnostic performance using bivariate or hierarchical summary ROC models. PROSPERO registration number: 2025 CRD420251058998.