Towards Explainable Artificial Intelligence for Pneumonia and Tuberculosis Classification from Chest X-Ray
Getamesay Haile, Meryam El Mouthadi
Abstract
The scientific community has shown significant interest in the application of deep learning models for the classification of tuberculosis and pneumonia from Chest X-ray (CXR) images. However, the primary emphasis of the majority of research lies only on enhancing classification accuracy. In this paper, we propose Explainable Artificial Intelligence (XAI) and a lightweight convolutional neural network (CNN) to enhance classification accuracy and explainability. The main contributions of this research are applying Contrast Limited Adaptive Histogram Equalization (CLAHE) to increase the visibility of CXR images, proposing a lightweight CNN, and evaluating it with a visual-based XAI model. We compare the proposed CNN with VGG16, Densnet201, EfficientNEtB0, InceptionV3, and MobileNetV2. We use the score-CAM XAI model to visualize why the model makes a certain decision. The proposed CNN has a classification accuracy of 97.5 <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup> , a precision of 97.4%, a 97.4% sensitivity, a 98.6% specificity, and a 97.4% F1 score on the testing set in multi-class classification. The findings indicate that deep learning and XAI can enhance trust in automatic disease detection and classification.