A Comparative Analysis of Chest X-ray Examination with AI Enhancement Using XAI Techniques

Abstract

Chest X-ray analysis plays a vital role in diagnosing pneumonia, and recent advancements in Deep Learning (DL) methods have significantly improved the accuracy of automated diagnosis. This study explores the intersection of DL and explainable artificial intelligence (XAI) in the context of pneumonia diagnosis through chest X-rays. The dataset used in this study consists of 1,341 training images of healthy individuals and 3,875 images of pneumonia cases, with the test set comprising 234 healthy and 390 pneumonia cases. Additionally, the validation set includes 8 images for both categories. This diversity aims to enhance the model's ability to generalize across different scenarios. The Convolutional Neural Network (CNN) and Transfer Learning (TL) methods utilizing the ResNet50 model achieved accuracies of 95.23 and 96.67, respectively. Subsequently, the models were explained using XAI methods such as SHAP and Grad-CAM. The study concludes by highlighting the potential of DL and XAI to enhance the interpretability and reliability of pneumonia diagnoses through chest X-ray analysis, aiming to contribute to future research in this field.

Ähnliche Arbeiten