Abstract
Accurate and timely diagnosis of meningitis is critical for effective clinical management, as delayed or incorrect identification can lead to severe neurological complications and increased mortality. Conventional diagnostic approaches rely heavily on cerebrospinal fluid analysis and laboratory investigations, which are often time-consuming. To overcome these limitations, this study proposes an optimized machine learning framework for meningitis diagnosis using a Multi-layer Perceptron neural network enhanced with an Immune-inspired Genetic Algorithm for hyperparameter optimization. The proposed model was evaluated on a structured clinical dataset comprising 5,925 patient records, incorporating demographic information, clinical symptoms, and key laboratory parameters such as CSF glucose, protein levels, leukocyte count, serum C-reactive protein, and procalcitonin levels. The dataset was categorized into three classes: bacterial meningitis, viral meningitis, and non-meningitis cases. Comprehensive preprocessing steps, including feature encoding, normalization, and data augmentation, were applied to enhance data quality and address class imbalance. The IIGA optimizes critical MLP hyperparameters by mimicking adaptive immune system mechanisms, including clonal selection, hypermutation, and affinity maturation, enabling effective exploration of the solution space. Experimental results demonstrate that the optimized MLP–IIGA model achieved a test accuracy of 99% and an unseen validation accuracy of 92%, outperforming the baseline MLP model, which achieved 98% and 90%, respectively. Additionally, notable improvements in precision, recall, and F1-score highlight the model’s enhanced robustness and generalization capability. The results confirm that immune-inspired evolutionary optimization significantly improves neural network performance, making the proposed framework a reliable, efficient, and promising clinical decision-support tool for meningitis diagnosis.
Keywords: Clinical Decision Support, Hyper Parameter Optimization, Immune-inspired Genetic Algorithm, Machine Learning, Meningitis, Multi-layer Perceptron.