Comparative Feature Extraction in KNN-Based Pneumonia Detection: PCA, HOG, SIFT, and LBP Analysis

Pneumonia remains a critical global health concern, underscoring the need for accurate, efficient diagnostic tools to support clinical decisions. This study evaluates the K-Nearest Neighbors (KNN) algorithm combined with three feature extraction methods—Principal Component Analysis (PCA), Histogram of Oriented Gradients (HOG), and Local Binary Patterns (LBP)—for automated pneumonia detection in chest X-ray images. Using a dataset with both normal and pneumonia cases, each feature extraction method was assessed across K values from 1 to 9. Experimental results showed that HOG with K=9 achieved the highest accuracy at 96.51% and an optimal AUC of 99.30%, followed by PCA with an accuracy of 95.28% and an AUC of 9832%. LBP demonstrated limited effectiveness, with lower accuracy and AUC scores compared to HOG and PCA. These findings highlight the robustness of HOG and PCA as reliable, low-cost diagnostic approaches for pneumonia detection, particularly in resource-constrained settings. Future work could explore additional machine learning algorithms, hybrid feature extraction approaches, and broader dataset diversity to enhance model generalizability and clinical applicability.