Meeting Banner
Abstract #1665

Fully Automated Intracranial Vessel Angiogram Segmentation from 4D flow MRI Data in Intracranial Stenosis Patients using Deep Learning 

Haben Berhane1, Maria Aristova2, Yue Ma2, Michael Markl2, and Susanne Schnell2
1Lurie Children's Hospital of Chicago, Chicago, IL, United States, 2Northwestern University, Chicago, IL, United States

We developed and validated a convolutional neural network for the fully automated 3D segmentation of the cerebral vasculature from 4D flow MRI for rapid flow analysis. Using 53 4D flow MRI scans, including 16 patients with intracranial atherosclerotic disease, we trained and tested our CNN using 10-fold cross validation. We assessed net flow and peak velocities across all of the major arteries and veins of the intracranial vasulature between automated and manually performed analysis. Across all metrics and regions, we found the automated segmentation showed excellent agreement with the manual, while taking a fraction of the time to perform.

This abstract and the presentation materials are available to 2020 meeting attendees and eLibrary customers only; a login is required.

Join Here