Please join us for a virtual Machine Learning Working Group Meeting on April 5, from 11 am-12 pm EST. Dr. Pingkun Yan, Assistant Professor in the Department of Biomedical Engineering at Rensselaer Polytechnic Institute (RPI) will present “Deep learning Predicts Risks from Chest CT”.
Abstract:
With the recent AI renaissance led by the advancement of deep learning, the landscape of medical imaging and analysis has undergone significant change. This talk presents a series of recent work on how deep learning helps enable new medical studies in chest computed tomography (CT). The first application tackles the challenge of predicting cardiovascular disease (CVD) risk from lung cancer screening low-dose CT. A developed deep neural network was trained with 30,286 LDCT volumes and achieved an area under the curve (AUC) of 0.869 on 2,085 National Lung Cancer Screening Trial subjects and identified patients with high CVD mortality risks (AUC of 0.768). The deep model further calibrated against the clinical gold standard CVD risk scores from ECG-gated cardiac CT. The promising results show the potential of converting LDCT for lung cancer screening into a dual-screening quantitative tool for CVD risk estimation. At the onset of the COVID-19 pandemic, we also developed image analysis techniques to predict ICU admission risk for COVID-19 patients from their chest CT scans and other non-imaging information with AUC up to 0.884. Altogether, these results encourage us to further develop powerful deep learning tools for pushing forward the frontiers of medical imaging and analysis.
The Center for Computational Imaging and Personalized Diagnostics has partnered with the Case Comprehensive Cancer Center to host the Machine Learning Working Group. Meetings are held with the express goal of leveraging the extensive Artificial Intelligence expertise concentrated across our partner institutions including Case Western Reserve University, University Hospitals, the Cleveland VA Medical Center, MetroHealth, and the Cleveland Clinic, and beyond. Through facilitating the interaction of basic and clinical researchers we hope to strengthen the scientific merit of presented studies as well as identify novel collaborative opportunities.
If you would like to attend this month’s session, please reach out to James Hale - jsh171@case.edu
