Probabilistic Models for Severity Assessment of Lung Disease

This project focuses on developing a series of interpretable probabilistic models to assess the severity of lung disease, particularly COVID-19 pneumonia, without sacrificing generality. Our model not only predicts severity class but also provides prediction uncertainty and saliency maps to enhance interpretability and reliability. This approach helps ensure better clinical understanding and trust in the predictions, using available data from chest X-rays and a multi-reader dataset.

Related Publications

• Zandehshahvar, M., van Assen, M., Kim, E., Kiarashi, Y., Keerthipati, V., Tessarin, G., Muscogiuri, E., Stillman, A.E., Filev, P., Davarpanah, A.H. and Berkowitz, E.A., 2024. Confidence-Aware Severity Assessment of Lung Disease from Chest X-Rays Using Deep Neural Network on a Multi-Reader Dataset. Journal of Imaging Informatics in Medicine, pp.1-11. Link
• Van Assen, M., Zandehshahvar, M., Maleki, H., Kiarashi, Y., Arleo, T., Stillman, A.E., Filev, P., Davarpanah, A.H., Berkowitz, E.A., Tigges, S. and Lee, S.J., 2022. COVID-19 Pneumonia Chest Radiographic Severity Score: Variability Assessment Among Experienced and In-Training Radiologists and Creation of a Multi-Reader Composite Score Database for Artificial Intelligence Algorithm Development. The British Journal of Radiology, 95(1134), p.20211028. Link
• Zandehshahvar, M., van Assen, M., Maleki, H., Kiarashi, Y., De Cecco, C.N. and Adibi, A., 2021. Toward Understanding COVID-19 Pneumonia: A Deep-Learning-Based Approach for Severity Analysis and Monitoring the Disease. Scientific Reports, 11(1), p.11112. Link