FXAI: fusing XAI for predicting COVID-19 using diverse chest x - ray images

Saleh, Radhwan A. A. and Farea, Shawqi Mohammed Othman and Al-Huda, Zaid and Ertunc, Metin and Kvak, Daniel and Al-antari, Mugahed A. (2023) FXAI: fusing XAI for predicting COVID-19 using diverse chest x - ray images. In: 18th International Conference on Intelligent Systems and Knowledge Engineering (ISKE), Fuzhou, China

Full text not available from this repository. (Request a copy)

Abstract

Fusing explainable artificial intelligence (FXAI) is currently a prominent research topic in medical imaging interpretations. The proposed FXAI has the capability to provide the following benefits. Firstly, it can extract strong and reliable high-level deep features by combining various standard AI networks. Secondly, it can simultaneously generate visual explainable saliency maps associated with each chest X-ray (CXR) scan. Such heat maps not only demonstrate the most relevant regions of the AI decision-making process but also offer advantages to radiologists and patients. Thirdly, it enhances prediction performance to deliver an optimal intelligent solution for communities worldwide. These advantages can support the development of an optimal treatment plan, reduce medical costs, and enhance the capabilities of health care systems. We have trained and evaluated the proposed FXAI using a diverse benchmark medical CXR dataset that has been collected from various public resources. Our findings encourage researchers and stakeholders in the medical industry to validate this proposed framework in a practical manner.
Item Type: Papers in Conference Proceedings
Uncontrolled Keywords: Deep Learning Classification; Diverse CO VID-I9 Dataset; Ensemble Learning; Fusing Explainable Artificial Intelligence (FXAI); XAI Saliency Maps
Divisions: Faculty of Engineering and Natural Sciences
Depositing User: Shawqi Mohammed Othman Farea
Date Deposited: 11 Jun 2024 20:57
Last Modified: 11 Jun 2024 20:57
URI: https://research.sabanciuniv.edu/id/eprint/49375

Actions (login required)

View Item
View Item