Texture-Based DNN for Pneumonia in Thorax X-Rays

This paper introduces an innovative methodology for identifying pneumonia in thoracic X-ray images through the application of neural network classifiers. In our experiment, we employed a comprehensive training regimen involving multiple neural network classifiers, each trained on distinct sets of texture features meticulously extracted from thoracic X-ray images. Four different gray-level matrices and a neighboring gray-tone difference matrix (NGTDM) were used to generate these input features, guaranteeing a reliable depiction of the textural properties found in the X-ray pictures. We carried out an extensive evaluation utilizing a number of performance criteria to gauge the trained classifiers' efficacy. Classifying the thoracic X-ray pictures into two groups’ pneumonia and healthy state was the assignment assigned to the classifiers. A thorough study of the classifiers' performance was provided by our assessment measures, which comprised accuracy, precision, recall, F1-score, and the area under the receiver operating characteristic curve (AUC-ROC). The experimental findings showed that the suggested method accomplished a remarkable 91% overall test categorization accuracy, which was encouraging. This degree of precision highlights how well our approach works to accurately diagnose pneumonia from thoracic X-ray images. Furthermore, the consistent performance across different metrics highlights the robustness and generalizability of the proposed strategy.