Lung infection classification using HOG based SVM classifier on X-ray images

Lung disease is one of the most prevalent reasons of death and affects many people in different ways globally. Pneumonia is a frequent respiratory illness brought on by bacteria or viruses. Chest X-rays, or CXRs, are a common method for detecting lung abnormalities, determining their severity, and recognizing possible issues. Multiple concurrent anomalies may be shown on chest X-rays. Using only the human eye to capture anomalies from a complex thoracic environment takes a lot of time, is delicate, and is prone to user prejudice. Radiologists must manually label patients, which uses a lot of medical resources. Because of overlapping radiographic patterns and inter-reader heterogeneity among radiologists, lung infections pose serious diagnostic problems. Traditional classification methods suffer from limited generalization across a variety of datasets, computational inefficiency, and difficulty extracting features. Autonomous techniques for detection connected to computer vision (CV) are now commonly employed in research fields including medical imaging. Utilizing both qualitative and quantitative medical data, machine learning (ML) algorithms have been found to be effective in disease prediction. In order to solve this problem, the current study suggests and uses Histogram Oriented Gradient (HOG) features that have been trained using a Support Vector Machine (SVM) classifier. Comparing HOG to other feature extraction methods, the outcomes implies that it provides superior structural characteristics to distinguish between two classes and well reflects the local description of edges. Additionally, based on the characteristics collected using the HOG technique, the SVM classifier is used to categorize the lung images that are infected and those that are not. According to experimental results, the suggested approach outperforms traditional methods like k-NN and decision trees in classification, with an accuracy of 94.23%. The model has issues with noise sensitivity and the need for high-quality picture preprocessing, notwithstanding its efficacy. Prompt evaluation and diagnosis can decrease the dangerous complications of lung disorders and improve the standard of life for those affected, as early detection and therapy are essential for the best results.