GRAPH NEURAL NETWORK (GNN)-BASED DECISION SUPPORT SYSTEM FOR LASSA FEVER DIAGNOSIS

The study presents a Graph Neural Network (GNN)-based Decision Support System for the diagnosis of Lassa fever. The system addresses the challenge of early detection due to symptom overlap with other febrile illnesses such as malaria and typhoid. A hybrid model combining Graph Convolutional Networks (GCN) and Graph Attention Networks (GAT) was developed to capture relationships between symptoms, patient features, and clinical data. Patient data, including symptoms and laboratory results, were modeled as a graph structure and processed using PyTorch Geometric. The system was evaluated using metrics such as accuracy, precision, recall, F1-score, and AUC-ROC. Results showed an accuracy of approximately 83.8% and an AUC of 0.705, demonstrating the effectiveness of relational learning in medical diagnosis. Despite promising performance, the model exhibited lower recall for positive cases, indicating the need for further optimization. The study concludes that GNNs offer a powerful approach for improving diagnostic systems, particularly in resource-constrained healthcare environments