1. An explainable transfer learning framework for multi-classification of lung diseases in chest X-rays
- Author
-
Aryan Nikul Patel, Ramalingam Murugan, Gautam Srivastava, Praveen Kumar Reddy Maddikunta, Gokul Yenduri, Thippa Reddy Gadekallu, and Rajeswari Chengoden
- Subjects
Computer-aided diagnosis systems ,Transfer learning ,EfficientNet-B4 architecture ,Explainable artificial intelligence ,Multi-disease lung classification ,Engineering (General). Civil engineering (General) ,TA1-2040 - Abstract
In the field of medical imaging, the increasing demand for advanced computer-aided diagnosis systems is crucial in radiography. Accurate identification of various diseases, such as COVID-19, pneumonia, tuberculosis, and pulmonary lung nodules, holds vital significance. Despite substantial progress in the medical field, a persistent research gap necessitates the development of models that excel in precision and provide transparency in decision-making processes. In order to address this issue, this work introduces an approach that utilizes transfer learning through the EfficientNet-B4 architecture, leveraging a pre-trained model to enhance the classification performance on a comprehensive dataset of lung X-rays. The integration of explainable artificial intelligence (XAI), specifically emphasizing Grad-CAM, contributes to model interpretability by providing insights into the neural network’s decision-making process, elucidating the salient features and activation regions influencing multi-disease classifications. The result is a robust multi-disease classification system achieving an impressive 96% accuracy, accompanied by visualizations highlighting critical regions in X-ray images. This investigation not only advances the progression of computer-aided diagnosis systems but also sets a pioneering benchmark for the development of dependable and transparent diagnostic models for lung disease identification.
- Published
- 2024
- Full Text
- View/download PDF