Benny Sukma Negara, Muhammad Irsyad, Laila Nurul Fauziyyah
This study presents a deep learning approach for classifying lung diseases from chest X-ray images using the VGG16 convolutional neural network combined with the Layer-wise Relevance Propagation (LRP) interpretability method. The dataset consists of 3,000 X-ray images in three classes (COVID-19, pneumonia, and normal), which are preprocessed through resizing, tensor conversion, normalization, and augmented with horizontal flipping and random rotation. Using transfer learning, VGG16 pretrained on ImageNet is adapted by freezing the feature extractor layers and retraining the final classification layer with a 70:30 train-test split, a learning rate of 0.001, batch size of 32, Adam optimizer, and CrossEntropyLoss. The trained model achieves high classification performance with 96.78% accuracy and balanced precision, recall, and F1-scores across classes. LRP is then applied to generate heatmaps highlighting image regions that most influence the model’s predictions, thereby increasing transparency and interpretability. The main contribution is the integration of VGG16 with LRP for multi-class chest X-ray classification, providing both accurate diagnostic support and visual explanations suitable for clinical use.
@article{156b08e3-9650-40eb-9505-cf31c7fc58cd,
title={Explainable Deep Learning for Lung Disease Detection from Chest X-rays via Layer-wise Relevance Propagation },
author={Benny Sukma Negara and Muhammad Irsyad and Laila Nurul Fauziyyah},
year={2026},
language={en}
}TY - JOUR TI - Explainable Deep Learning for Lung Disease Detection from Chest X-rays via Layer-wise Relevance Propagation AU - Benny Sukma Negara AU - Muhammad Irsyad AU - Laila Nurul Fauziyyah PY - 2026 LA - en ER -
Roger Rumbu
This paper provides a detailed investigation into the role of hot gas circulation in metallurgical fluid bed roasters, aiming to enhance heat and mass