Machine Learning for Lung Cancer Screening | July 7, 2020
Faculty Speaker: Dr. Vineet Raghu, Lecturer
Moderated by: Andrey Blidman, Computer Science Program Coordinator
July 7, 2020
Abstract: Lung cancer is the leading cause of cancer death in the United States, as it accounts for more cancers than the next three cancer types combined. The primary known cause of lung cancer is cigarette smoking; however, it is unclear how to decide which smokers are at the highest risk for developing lung cancer. The National Lung Screening Trial showed that mortality due to lung cancer can be reduced through the use of screening. Screening consists of scanning the chest using low-dose Computed Tomography to search for the presence of abnormal growths in the lung (lung nodules). However, the majority of lung nodules (96%) turn out to be benign and do not become cancerous. This raises two pressing questions: 1) How do we determine who should be screened for lung cancer? and 2) How do we know which nodules will eventually become cancerous? In this talk, I explore whether machine learning techniques can be used to address this question. I discuss deep learning from chest x-rays to select high-risk individuals for screening, and I discuss graphical models from CT scan data to identify lung nodules likely to become cancerous.