A Comparative Analysis of Deep Learning Models for Detecting Malaria Disease Through LBP Features

Malaria is a parasitic infection spread by the plasmodium parasite. Malaria continues to be a major threat to world health, with an estimated 200 million cases and over 400,000 fatalities each year. When exposed to this disease, symptoms develop 10–15 days after the parasite enters the body. This disease becomes chronic if it is not treated medically, and it eventually leads to death. Using spatial information collected from microscopic images, several techniques based on image processing and machine learning have been utilized to diagnose malaria. Using the Local Binary Pattern (LBP) texture feature as a feature extraction approach, this study contributes to the development of a predictive and high-accuracy deep learning model by testing multiple Deep Learning models and determining which model delivers the best accuracy. To be specific, we tested frequently used baseline methods, namely ResNet34, VGG16, Inception V3, and EfficientNet. The results demonstrate that EfficientNet has a 91 percent outstanding accuracy rate, compared to 87 percent for VGG16, 81 percent for Resnet34, and 77 percent for InceptionV3, respectively.