ICYMI: AI + Health Seminar with Edward Castillo
Title: Predicting Long Term Mortality in COPD Using Deep Learning Imaging Markers
Speaker: Edward Castillo, PhD, Associate Professor & Associate Chair for Graduate Education, William J. Murray, Jr. Fellowship in Engineering, Department of Biomedical Engineering, University of Texas at Austin
Date: October 23, 2025
Presentation: Zoom link
Abstract: Chronic obstructive pulmonary disease (COPD) remains a major global health challenge, and current clinical tools often fall short in forecasting long-term outcomes. We present new modeling approach that leverages deep learning to extract imaging biomarkers from paired inspiratory and expiratory CT scans. Using data from over 8,800 participants in the COPDGene study, our research team developed a fused model that integrates these imaging features with clinical variables, including BMI, obstruction, dyspnea, and exercise capacity. Out model achieves a significant improvement in 10-year mortality prediction, with a concordance index of 0.78—outperforming traditional methods. This work highlights the complementary value of multi-phase CT imaging and AI-driven analysis to support a more individualized approach to prognosis and disease management, offering clinicians additional tools to better tailor care for patients with COPD.
Speaker Bio: Edward Castillo, PhD is an Associate Professor and the Associate Chair for Graduate Education in the Biomedical Engineering Department at UT Austin. He directs the Dynamic Medical Image and Computing Lab, where his research focuses on developing computational methods for medical image analysis, disease progression modeling, and improving radiotherapy outcomes. Dr. Castillo has held faculty positions at MD Anderson Cancer Center and Oakland University William Beaumont School of Medicine, and is a Charter Member of the NIH Image-Guided Intervention and Surgery Study Section
Title: Predicting Long Term Mortality in COPD Using Deep Learning Imaging Markers
Speaker: Edward Castillo, PhD, Associate Professor & Associate Chair for Graduate Education, William J. Murray, Jr. Fellowship in Engineering, Department of Biomedical Engineering, University of Texas at Austin
Date: October 23, 2025
Presentation: Zoom link
Abstract: Chronic obstructive pulmonary disease (COPD) remains a major global health challenge, and current clinical tools often fall short in forecasting long-term outcomes. We present new modeling approach that leverages deep learning to extract imaging biomarkers from paired inspiratory and expiratory CT scans. Using data from over 8,800 participants in the COPDGene study, our research team developed a fused model that integrates these imaging features with clinical variables, including BMI, obstruction, dyspnea, and exercise capacity. Out model achieves a significant improvement in 10-year mortality prediction, with a concordance index of 0.78—outperforming traditional methods. This work highlights the complementary value of multi-phase CT imaging and AI-driven analysis to support a more individualized approach to prognosis and disease management, offering clinicians additional tools to better tailor care for patients with COPD.
Speaker Bio: Edward Castillo, PhD is an Associate Professor and the Associate Chair for Graduate Education in the Biomedical Engineering Department at UT Austin. He directs the Dynamic Medical Image and Computing Lab, where his research focuses on developing computational methods for medical image analysis, disease progression modeling, and improving radiotherapy outcomes. Dr. Castillo has held faculty positions at MD Anderson Cancer Center and Oakland University William Beaumont School of Medicine, and is a Charter Member of the NIH Image-Guided Intervention and Surgery Study Section