AI-Enabled MR Reconstruction: Doing More With Less
, Senior Applications Engineer, GE Healthcare
Compared to other imaging modalities, magnetic resonance (MR) imaging scans have been historically challenged with relatively long data acquisition times, requiring patients to remain still, or for some procedures, hold their breath. With the advent of multi-channel RF coils, MR scans could be modestly accelerated (2–4x) using parallel imaging techniques. We'll cover the development of an AI-based reconstruction solution — Sonic DL Cine — that offers significantly higher reductions in scan time while retaining diagnostic image quality. The application is currently supported for cardiac imaging and provides images of the heart in multiple cardiac phases from systole to diastole — within a single heartbeat. We'll cover the design considerations made during the training, testing, and deployment of the final product, all of which employed NVIDIA’s software and hardware infrastructure. While some knowledge of MR imaging will be useful, it isn't essential.
Events & Trainings:
Date:
Industry:
Level:
Topic:
NVIDIA technology: Cloud / Data Center GPU,CUDA,Triton