Multi-Class Classification of Chest X-ray Images with Optimized Features

A global health challenge has been created since the pandemic of COVID-19. The widespread growth of diseases like pneumonia, and lung cancer as side effects of COVID-19. The rapid and accurate identification of these diseases becomes crucial to hinder their spread and start early treatments where Machine Learning (ML) based computer automated diagnosis performs best. The comprehensive approach here integrates preprocessing, feature extraction using Local binary pattern (LBP), histogram of oriented gradients (HOG), and transfer learning using EfficientNet-B0 models. Then the extracted features are optimized with a Bayesian optimizer and fed to SVM and Random Forest classifier for classification. Thorough evaluation and hyperparameter optimization play an essential role in the effective classification of this approach. The SVM classifier was found to be performing better in comparison to the Random forest classifier in terms of metrics like precision, recall and F1-score with an overall accuracy of 97%.