Penerapan Model SVM dengan Ekstraksi Fitur ResNet50 untuk Identifikasi Sel Darah Terinfeksi Malaria

Adhesyah Putra, Maulana Damar
Rakasiwi, Sindhu

Building of Informatics, Technology and Science
Vol 7 No 3 (2025): December 2025

Malaria remains a major public health challenge in Indonesia, with 279,865 reported cases in 2023 and an Annual Parasite Incidence (API) of 0.99 per 1,000 population. Although microscopic examination is still considered the gold standard for malaria diagnosis, it has several limitations, including dependency on trained experts, subjective interpretation, and relatively lengthy processing time. To address these challenges, this study aims to analyze the performance of a Support Vector Machine (SVM) classifier with feature extraction based on ResNet50 in a Computer-Aided Diagnosis (CAD) system for automatic detection of malaria-infected blood cells.ResNet50 was selected for its transfer learning capability to generate high-level feature representations from medical images, while SVM was chosen due to its strong performance on high-dimensional data and limited datasets. A feature vector of 2048 dimensions produced from the global average pooling layer was classified using SVM with a Radial Basis Function (RBF) kernel. The dataset used in this study was obtained from the National Institutes of Health (NIH) and consists of 27,558 microscopic blood cell images (Parasitized and Uninfected classes). The data were partitioned using stratified sampling with an 80:20 ratio for training and testing. Preprocessing steps included pixel normalization, resizing to 224×224 pixels, and basic augmentation to improve model generalization. Experimental results show that the proposed model achieved an accuracy of 93.94%, precision of 94%, recall of 93.43% (Parasitized) and 94.46% (Uninfected), and an average F1-score of 94%. The confusion matrix indicates 2,575 true positives, 2,606 true negatives, 153 false positives, and 181 false negatives, with a false negative rate of 6.57% and a false positive rate of 5.54%. These findings demonstrate that the combination of ResNet50 and SVM has strong potential as a fast and accurate image-based malaria detection method and is suitable for implementation in healthcare settings with limited resources.