Comparative Analysis of Deep Learning Models for Chest X-ray Classification and Interpretability using Explainable AI Techniques

Medical Image analysis is a crucial aspect of modern healthcare, and deep learning models have shown promising results in this area. However, the interpretability of these models remains a challenge, particularly for complex medical images such as chest X-rays. In this project, we evaluate the performance of three state-of-the-art deep learning models, Inceptionv4, Vision Transformers, and ResNet50, for the classification of chest X-ray images using two datasets, NIH-CXR-LT and MIMIC-CXR-LT. Our findings show the comparison of these models with AUC scores on the chest X-ray datasets, with Vision Transformers having a better AUC score and reduced training time. Furthermore, we demonstrate the interpretability of the models using explainable AI techniques. This project has the potential to improve the accuracy and interpretability of deep learning models for medical image analysis, which could have significant implications for the future of healthcare.