Region-specific cardiovascular risk prediction from non-contrast chest computed tomography

Vineet K Raghu, Daniel W Oo, Leonard Nürnberg, Audra Sturniolo, Douglas P Kiel, Hugo Aerts, Pradeep Natarajan, Michael T Lu

Keywords: CT imaging, cardiovascular risk prediction, convolutional networks, segmentation

TL;DR: Convolutional networks can estimate cardiovascular risk from non-contrast chest CT by first extracting relevant anatomic regions of interest

Abstract: Accurate cardiovascular risk scores can help direct preventive treatment to those who would maximally benefit. Current scores rely on established risk factors, but imaging may contain additional information to find high-risk patients. Here, we developed and tested a system called CT-CV-Risk to estimate cardiovascular risk from non-contrast chest CT images. We find that CT-CV-Risk predicts risk complementary to established clinical risk scores.

Submission Number: 92