Abstract: A promising approach to portable MRI is operation at ultra-low magnetic field where cost-effective electromagnets become practical. MRI in the ultra-low field (ULF) regime—when the magnetic field used for signal detection is below 10 mT—is inherently challenging mainly due to intrinsically low Boltzmann polarization. We will discuss hardware methods to improve attainable SNR in the Johnston-noise-dominated regime of ULF using improved coils such as quadrature volume coils at 276 kHz (6.5 mT). We will also discuss our work to reduce noise and increase attainable information per unit time using compute-based approaches that leverage low-cost GPU. In particular, we focus this talk on our neural network deep learning approach, AUTOMAP (AUtomated TransfOrm by Manifold APproximation), to reconstruct highly-undersampled low SNR imaging data using an approach inspired by perceptual learning that occurs in biological visual systems.

Bio: Dr. Matt Rosen is a physicist, tool-builder, and inventor whose research bridges the spectrum from fundamental physics to applied bioimaging work in the field of MRI. He established the Low-Field MRI and Hyperpolarized Media Laboratory at the Athinoula A. Martinos Center for Biomedical Imaging to focus on the continued development of new hyperpolarization methods and MRI-based tools. The Rosen Lab focuses on new methods and tools to enable unconventional approaches to MRI scanner construction. This includes the development of new acquisition strategies for robust ultra-low magnetic field implementations of MRI focused on brain imaging. The laboratory also explores opportunities provided by hyperpolarization including in vivo Overhauser DNP, SABRE, and spin-exchange optical pumping. We create new quantitative strategies for the acquisition and the reconstruction of highly-undersampled imaging data including neural network deep learning based approaches such as AUTOMAP that leverage low-cost scalable-compute. Matt co-directs the Center for Machine Learning at Martinos.

BME Faculty Hosts: Joseph Rispoli and Uzay Emir