Classification of Malaria Parasite Images Using Gray Level Co-Occurrence Matrix Feature Extraction and Neural Networks

Malaria is a life-threatening infectious disease. In recent years, Plasmodium knowlesi (PK) has emerged as a significant concern in malaria cases, particularly in Southeast Asia, including countries such as Indonesia, Malaysia, and Thailand. One of the most commonly adopted methods is microscopic examination of stained blood smears, such as Giemsa-stained smears. However, manual counting and classification of cells is time-consuming. Automated detection systems using machine learning image processing techniques have demonstrated promise in addressing these challenges. This study combined two Multilayer Perceptron (MLP) optimizations with Gray Level Co-occurrence Matrix (GLCM) feature extraction. The optimization methods were Broyden Fletcher Goldfarb Shanno (BFGS) and Levenberg Marquardt (LM). The results illustrated that LM optimization consistently outperformed BFGS in terms of both training and testing accuracy across all configurations, achieving greater stability and generalization on unseen data. LM optimization achieved the highest average accuracy of 94.80% for training and 79.76% for testing. It revealed a significant improvement over BFGS models, with training times averaging 3.90 seconds, reflecting an efficient balance between accuracy and computational demand.