DEVELOPMENT OF HYBRID ARTIFICIAL INTELLIGENCE-BASED MEDICAL DIAGNOSTIC SYSTEM FOR PREDICTION AND EARLY DETECTION OF LYMPHATIC FILARIASIS USING CLINICAL DATA

Lymphatic filariasis remains one of the most persistent neglected tropical diseases in endemic regions, with diagnosis still largely dependent on manual microscopic examination a process constrained by the scarcity of trained microscopists and susceptible to observer variability. Existing artificial intelligence solutions have attempted to close this gap, but most rely on a single data stream, typically image classification alone, leaving out clinical context that experienced physicians would ordinarily draw on. This study proposes a medical intelligence system that fuses microscopic blood smear analysis with patient clinical data (demographic profile, symptoms, and laboratory findings) to detect and predict lymphatic filariasis across four diagnostic categories: infected, non-infected, early-stage, and late-stage. The system was built using a hybrid of Design Science Research Methodology and Object-Oriented Analysis and Design Methodology, with visual features extracted through a pretrained VGG16 network and structured clinical variables incorporated through a feature-fusion process. Four algorithms Logistic Regression, Random Forest, XGBoost, and VGG16 were trained and benchmarked against one another, with VGG16 emerging as the strongest performer at 94% validation accuracy and an F1-score of 0.93. Class-wise evaluation showed the model distinguished non-infected and infected cases with particular consistency, while early-stage and latestage classifications, though slightly less precise due to overlapping diseaseprogression features, remained clinically usable. Beyond raw accuracy, the results suggest a practical shift: diagnostic support that does not require an on-site expert microscopist could extend earlier detection to primary healthcare facilities in low-resource, high-burden regions. The paper concludes that multimodal fusion offers a measurable advantage over image-only diagnostic pipelines and outlines directions for future work, including explainable AI integration and mobile/cloud deployment for realtime field use