Automated COVID-19 detection from chest X-rays using HOG features

The COVID-19 pandemic has necessitated the development of rapid and accurate diagnostic tools to assist healthcare professionals in disease detection and management [1]. This study presents a different machine learning framework for COVID-19 classification using chest X-ray images, employing Histogram of Oriented Gradients (HOG) feature extraction combined with Principal Component Analysis (PCA) for dimensionality reduction and ensemble learning methods for robust classification. The methodology integrates seven different classifiers including Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Random Forest, Decision Tree, Naive Bayes, Linear Discriminant Analysis (LDA), and an ensemble voting classifier. Image preprocessing techniques including adaptive histogram equalization and HOG feature extraction with 8×8 cell size were applied to 128×128 resized chest X-ray images. PCA was utilized to reduce the feature dimensionality while preserving essential discriminative information, with the first 100 principal components retained for classification. The ensemble approach combined predictions from SVM, KNN, and Random Forest classifiers using majority voting to enhance diagnostic accuracy. Results were validated through 10 independent runs to ensure statistical reliability and reduce model variance. The proposed framework demonstrates the effectiveness of traditional computer vision techniques combined with machine learning algorithms for medical image analysis, providing a computationally efficient alternative to deep learning approaches while maintaining competitive performance in COVID-19 detection from radiological images.