Efficient MRI Through Improved Encoding & Reconstruction
In this talk, we will introduce encoding and reconstruction strategies that aim to “fix” echo planar imaging by eliminating geometric distortions, boosting acquisition efficiency and reducing T2/T2* related voxel blurring. These strategies allow for rapid and high resolution acquisitions with isotropic voxel sizes and lend themselves to diffusion imaging, relaxation parameter estimation and quantitative susceptibility mapping.
We show that these approaches can be also tailored to provide a 2 minute comprehensive brain exam with multiple clinical contrasts with high geometric fidelity and SNR. Finally, we introduce physics-driven machine learning reconstruction techniques that further improve the achievable acceleration rates and image quality. These incorporate what is known about the encoded image without being exposed to potential vulnerabilities of deep learning in terms of generalizability and explainability. Tailoring these models so that they can work in a scan-specific, unsupervised manner further addresses concerns about generalization, as they are trained solely using acquired data per each individual subject.
About the Speaker
Berkin Bilgic holds degrees in Physics and Electrical Engineering (BS) and Electrical Engineering and Computer Science (MS, PhD). He is an Assistant Professor in Radiology at the Martinos Center for Biomedical Imaging, Massachusetts General Hospital, and Harvard Medical School, where his group develops data acquisition and reconstruction techniques for MRI.
Join Zoom Meeting
https://illinois.zoom.us/j/81269558758?pwd=SXF4K1dyZkRUL0ZHOHlLUjZlMGc2dz09
Meeting ID: 812 6955 8758
Password: 585668
One tap mobile
+13126266799,,81269558758# US (Chicago)
+14702509358,,81269558758# US (Atlanta)